Linux isn't ready for the enterprise, according to M. Russinovich from sysinternals.com .
Boris
Marks claims that Linux 2.2 doesn't scale sufficiently well.
This makes sense: OS intended for poor students with 386 cannot be suitable for the
enterprise.
From my personal experience with several ISPs freeware UNIXes aren't ready for WEB hosting
either.
Boris
Jeremy Allison
So why does Samba running on Linux on the same
hardware beat NT performance once you go above 8 clients ?
How come NT doesn't scale better than Linux based on
the theoretical arguments in the NT magazine article ?
(See http://www.zdnet.com/sr/stories/issue/0,4537,396321,00.html
for details).
Note that this isn't a theoretical "select is really bad
'cos Microsoft say so" article like the NT mag one (I remember
them peddling that story at the '94 Developers conference,
funny how some things don't change) but a real test done
on real hardware as shipped by real vendors.
Care to explain these results Boris ?
Regards,
Jeremy Allison.
Stephen Edwards
: The latest NT Magazine: April 99.
: Marks claims that Linux 2.2 doesn't scale sufficiently well.
IIRC, Linux can only scale to 16 processors. WindowsNT can scale to
twice that many... Yet "WindowsNT isn't scalable" in the eyes of most
Linux users... silly.
: This makes sense: OS intended for poor students with 386 cannot be
: suitable for the enterprise.
Linux does pretty well for your basic mail and news services. I think
what's keeping most corporations away from Linux is the fact that it
is so easily exploitable.
: From my personal experience with several ISPs freeware UNIXes aren't
: ready for WEB hosting either.
Any halfway decent ISP would use BSD/OS over Linux any day.
[] "No footnote for you!" -- Footnote Nazi
--
.-----.
|[_] :| Stephen S. Edwards II | http://www.primenet.com/~rakmount
| = :| "But something's wrong. It takes me a moment to pin it down;
| | the monitor's different. Instead of the nice 17' Trinitron,
|_..._| there's a 15' raster gun in a dirty plastic case." -- Ben in ASR
Boris
are mostly gone.
Plus ZD test machines in NT vs. Linux results you describe had only 64MB of RAM; this is
clearly insufficient amount of RAM for production NT server. To make my point clearer: DOS
beats NT with <16 MB of RAM (because NT cannot even run with less than 16 MB). While there
are dozens of NT benchmark results on www.tpc.org and www.spec.org there isn't a single
mentioning of Linux on those sites. Which just proves the thing: nobody uses this thing
(Linux) as enterprise server platform as opposed to NT (which is used very widely). Mark
gave theoretical explanation of the fact that Linux isn't usable in enterprise server
role.
Boris
Jeremy Allison wrote in message ...
>"Boris" <boris...@pleasemovil.com> writes:
>
>>The latest NT Magazine: April 99.
>>Marks claims that Linux 2.2 doesn't scale sufficiently well.
>>This makes sense: OS intended for poor students with 386 cannot be suitable for the
>>enterprise.
>>From my personal experience with several ISPs freeware UNIXes aren't ready for WEB
hosting
>>either.
>
Jerry
Can you define what you mean by an "enterprise" server ? Presumably non
of the majority of ISPs who use Linux/Apache are counted as
"enterprises" ?
Seems to me you are talking out of your bottom.....
Cheers,
Jerry
mlw
>
> Boris <XXXb...@movil.comXXX> wrote:
>
> : The latest NT Magazine: April 99.
> : Marks claims that Linux 2.2 doesn't scale sufficiently well.
>
> twice that many... Yet "WindowsNT isn't scalable" in the eyes of most
> Linux users... silly.
What released version of NT can scale to more that 4 processors?
Also, NT clustering is a joke, it is primarily a fail-over cluster.
Linux on the other hand has can do hundreds of computers and a Linux
cluster has been rated as one of the fastest computers in the world.
--
Mohawk Software
Windows 95, Windows NT, UNIX, Linux. Applications, drivers, support.
Take the Mohawk Software Computer Survey at: www.mohawksoft.com
Darren Winsper
> Boris <XXXb...@movil.comXXX> wrote:
>
> : The latest NT Magazine: April 99.
> : Marks claims that Linux 2.2 doesn't scale sufficiently well.
>
> IIRC, Linux can only scale to 16 processors. WindowsNT can scale to
> twice that many... Yet "WindowsNT isn't scalable" in the eyes of most
> Linux users... silly.
OK, NT does not scale effectively beyond 4 CPUs. Just because NT can
handle more CPUs doesn't mean it can scale effectively.
--
Darren Winsper
"Microsoft says, "Oh, you've got a brain? Well, you won't need it as
long as you stay within this nice little space we've prepared for you."
Linux says, "Oh, you've got a brain? Splendid! Here are lots of fun
things to do with it."" - Daniel Birchall
bill
Boris wrote in message <370adca9$0$2...@nntp1.ba.best.com>...
>The latest NT Magazine: April 99.
>Marks claims that Linux 2.2 doesn't scale sufficiently well.
the
>enterprise.
>From my personal experience with several ISPs freeware UNIXes aren't ready for
WEB hosting
>either.
>
That's a laugh. For an example of a Linux/Apache site, try MP3.com. I'd be
willing to bet that their daily throughput, as measured in bytes served, is in
the top 50 sites worldwide (probably top 10, excluding porn). Linux has already
won... it's just a matter time.
Bill House
>Boris
>
>
Stephen Edwards
: Stephen Edwards wrote:
: >
: > Boris <XXXb...@movil.comXXX> wrote:
: >
: > : Marks claims that Linux 2.2 doesn't scale sufficiently well.
: > IIRC, Linux can only scale to 16 processors. WindowsNT can scale to
: > twice that many... Yet "WindowsNT isn't scalable" in the eyes of most
: > Linux users... silly.
: What released version of NT can scale to more that 4 processors?
All an OEM needs to do, is make minor adjustments to the HAL. WindowsNT,
according to NTFAQ.com, can support up to 32 processors if the proper
changes to the HAL are made. If you're referring to "off the shelf"
WindowsNT, then no, you're limited to 4 processors, unless you're willing
to deal with an OEM. That fact does not invalidate the fact that
WindowsNT can support more processors than Linux.
All I'm saying is that the statement "WindowsNT is not scalable" is false.
: Also, NT clustering is a joke, it is primarily a fail-over cluster.
I have also heard that clustering under WindowsNT v4.0 is less than
satisfactory. But so what? Progress takes time. If someone really needs
clustering, they'll use some release of UNIX, until Microsoft releases an
OS which can handle it properly.
: Linux on the other hand has can do hundreds of computers and a Linux
: cluster has been rated as one of the fastest computers in the world.
But Linux isn't the only OS with this capability. Most commercial UNIX
variants can do this sort of thing very well.
Jeremy Allison
>NT is optimized as application server (as opposed to file server). Days of file servers
>are mostly gone.
>Plus ZD test machines in NT vs. Linux results you describe had only 64MB of RAM; this is
>clearly insufficient amount of RAM for production NT server. To make my point clearer: DOS
>beats NT with <16 MB of RAM (because NT cannot even run with less than 16 MB). While there
>are dozens of NT benchmark results on www.tpc.org and www.spec.org there isn't a single
>mentioning of Linux on those sites. Which just proves the thing: nobody uses this thing
>(Linux) as enterprise server platform as opposed to NT (which is used very widely). Mark
>gave theoretical explanation of the fact that Linux isn't usable in enterprise server
>role.
Boris, I worked with PC Week on their Linux 2.2 + Samba
benchmark on a VA Research machjine also. That box had 4
400 MHz Xeon processors and 2 GB of memory with 5 disks
in a raid 5 configuration.
The NetBench results for NT serving NT clients were *so
bad* that PC Week didn't want to publish them. The reason
it turned out was that NT *will not* use more than 300Mb
of memory as a file system cache. I was able to tune Linux
to use 1.8Gb as file system cache.
I love the "days of file server only are mostly gone"
comment. Isn't that a "well we cannot win in that
area so lets pretend it doesn't count" comment :-) :-).
What are you going to say when Oracle publishes
benchmark numbers on Linux. Will it be "days of
database server are mostly gone.....".
Most NT implementations in my experience are
used as small file and print servers. If it isn't
much good for that, what *is* it good for ? Obviously
not Web serving, Apache holds <50% market share
there....
So tell me again how "enterprise ready" NT is ? But only
if you don't have more that 512Mb of Ram maybe ?
BTW: The 300Mb limit is actually *documented* in the
Windows NT Internals book (2nd edition).
Regards,
Jeremy Allison.
Joseph E Mainusch
>
> The latest NT Magazine: April 99.
> Marks claims that Linux 2.2 doesn't scale sufficiently well.
> enterprise.
> From my personal experience with several ISPs freeware UNIXes aren't ready for WEB hosting
> either.
>
Linux doesn't scale well.
NT scales beautifully.
Hmmm... Then why are there no NT machines in the TOP500 list of
supercomputers (http://www.top500.org/), while several Linux machines
are?
--
Joseph E. Mainusch
65 Jerome Ave
Mineola, NY 11501-3301
(516)742-3874
David J. Owens
>
> That's a laugh. For an example of a Linux/Apache site, try MP3.com. I'd be
> willing to bet that their daily throughput, as measured in bytes served, is in
> the top 50 sites worldwide (probably top 10, excluding porn). Linux has already
> won... it's just a matter time.
According to Netcraft's Web Server Survey(http://www.netcraft.com/cgi-bin/Survey/whats),
mp3.com is running on FreeBSD-Apache, not Linux-Apache. Netcraft
does get incorrect statistics for some web sites, however.
Stephen Edwards
: "Boris" <boris...@pleasemovil.com> writes:
[snip]
: >gave theoretical explanation of the fact that Linux isn't usable in enterprise server
: >role.
: The NetBench results for NT serving NT clients were *so
: bad* that PC Week didn't want to publish them. The reason
: it turned out was that NT *will not* use more than 300Mb
: of memory as a file system cache. I was able to tune Linux
: to use 1.8Gb as file system cache.
My main problem with benchmarks of any sort are "was each operating system
configured optimally, and fairly?" I'm not accusing anyone of lying, but
I do wonder if most of these tests are fair. It's quite clear that many
UNIX experts are rather ignorant about WindowsNT, as many WindowsNT
experts are equally ignorant about UNIX.
How are we to be certain that these tests are done fairly, Jeremy?
: Most NT implementations in my experience are
: used as small file and print servers. If it isn't
: much good for that, what *is* it good for ? Obviously
: not Web serving, Apache holds <50% market share
: there....
I agree that Apache is quite a nice piece of software, but let's be
realistic here... partly why Apache has taken hold so easily is it's cost.
Who wants to shell out 1,000's of $ for UNIX Web server software, when
there is an excellent solution for free?
Mind you, some would say the same thing about UNIX in general, but let's
also not forget that there are some things commercial UNIX have that
OpenSource(tm) UNIXen do not.
: So tell me again how "enterprise ready" NT is ? But only
: if you don't have more that 512Mb of Ram maybe ?
: BTW: The 300Mb limit is actually *documented* in the
: Windows NT Internals book (2nd edition).
Which version of WindowsNT does this book refer to? WindowsNT v4.0 has
had major changes made to it from v3.5.
Shimpei Yamashita
>
>Boris <XXXb...@movil.comXXX> wrote:
>
>: The latest NT Magazine: April 99.
>: Marks claims that Linux 2.2 doesn't scale sufficiently well.
>
>twice that many... Yet "WindowsNT isn't scalable" in the eyes of most
>Linux users... silly.
OTOH, NT doesn't make good use of anything above four microprocessors.
Linux users who care about scalability are probably also aware of
"real" scalable OS's like Solaris, which not only goes up to 64
processors but can take advantage of all of them.
>
>: This makes sense: OS intended for poor students with 386 cannot be
>: suitable for the enterprise.
>
>Linux does pretty well for your basic mail and news services. I think
>what's keeping most corporations away from Linux is the fact that it
>is so easily exploitable.
And then they adopt an operating system that uses ActiveX, vulnerable
to Word macro viruses like--erm--Melissa, has a halfhearted
implementation of multiuser environment, etc., etc. Right. Makes a lot
of sense.
Linux and other Freeware Unices *look* like they have security
problems because all its problems get publicized right away. However,
the flip side of the equation is that their holes get *fixed* the
moment they are found. The proprietary OS's *look* like they have
no security problems because they don't get publicized. That doesn't
mean they don't exist; Solaris 2.6 probably requires at least as many
_published_ security patches as a RedHat 5.x installation does.
I don't have much experience with NT, but NT does get its share of
coverage on BugTraq.
Fundamentally, the most important part of a secure OS is the
administrator. Any OS can be exploited in the hands of a lazy or
incompetent administrator. I do not think most Linux distributions are
no more or less secure out of the box than commercial Unices.
>: From my personal experience with several ISPs freeware UNIXes aren't
>: ready for WEB hosting either.
>
>Any halfway decent ISP would use BSD/OS over Linux any day.
ftp.cdrom.com is using FreeBSD over proprietary Unix. Dejanews is using
Linux. Yahoo (the last I heard) uses a combination of the two. Do these
qualify as "halfway decent ISP"s?
And from *my* personal experience, Solaris was just as problematic at
web hosting as Linux would have been, albeit in different ways. Just
what problem does Linux have that commercial Unices don't that makes
them not suitable for web hosting? The only problem I can see is
hardware: most PCs just aren't ready for serious 24/7 server duty. But
that isn't Linux's fault.
--
Shimpei Yamashita <http://www.submm.caltech.edu/%7Eshimpei/>
Chris Costello
> The latest NT Magazine: April 99.
> Marks claims that Linux 2.2 doesn't scale sufficiently well.
> enterprise.
> either.
Freeware? First off, Linux isn't Unix. Second off, Freeware
is a DOS/Windows user term for software that costs $0.
>
> Boris
>
>
--
=====================================================
* "x86 ASM is for those who enjoy a bullet to the *
* foot from time to time. The occasional anti- *
* aircraft missile, too." - Chris Costello *
=====================================================
Craig Kelley
[snip]
> I agree that Apache is quite a nice piece of software, but let's be
> realistic here... partly why Apache has taken hold so easily is it's cost.
> Who wants to shell out 1,000's of $ for UNIX Web server software, when
> there is an excellent solution for free?
We actually use Apache on two of our NT boxes, instead of IIS (which
is also "free"). Admittedly, this was because we already knew how to
get perl to work with it. It was nice to simply be able to copy our
entire web database system from the AIX box to the NT box and have it
all run flawlessly. (that, and the new SMC was more difficult to
understand than httpd.conf)
I imagine we'd have had problems if we had wanted to go the other way,
using ASP.
> : If you don't have more that 512Mb of Ram maybe ?
>
> : BTW: The 300Mb limit is actually *documented* in the
> : Windows NT Internals book (2nd edition).
>
> Which version of WindowsNT does this book refer to? WindowsNT v4.0 has
> had major changes made to it from v3.5.
The study which he is talking about was done not more than four
months ago (this was after NT4/SP4 was out).
--
The wheel is turning but the hamster is dead.
Craig Kelley -- kell...@isu.edu
http://www.isu.edu/~kellcrai finger i...@inconnu.isu.edu for PGP block
Stephen Edwards
: Stephen Edwards <ja20...@primenet.com> writes:
: [snip]
: > I agree that Apache is quite a nice piece of software, but let's be
: > realistic here... partly why Apache has taken hold so easily is it's cost.
: > Who wants to shell out 1,000's of $ for UNIX Web server software, when
: > there is an excellent solution for free?
: We actually use Apache on two of our NT boxes, instead of IIS (which
: is also "free"). Admittedly, this was because we already knew how to
: get perl to work with it. It was nice to simply be able to copy our
: entire web database system from the AIX box to the NT box and have it
: all run flawlessly. (that, and the new SMC was more difficult to
: understand than httpd.conf)
How does Apache stand up on NT, performance-wise?
: I imagine we'd have had problems if we had wanted to go the other way,
: using ASP.
: > Which version of WindowsNT does this book refer to? WindowsNT v4.0 has
: > had major changes made to it from v3.5.
: The study which he is talking about was done not more than four
: months ago (this was after NT4/SP4 was out).
I see. It could be that Microsoft was more concerned with memory
avaialble to productivity applications than they were memory for other
tasks.
Either way, it definitely is a hinderance.
mlw
> : >
> : > : The latest NT Magazine: April 99.
> : > : Marks claims that Linux 2.2 doesn't scale sufficiently well.
> : > IIRC, Linux can only scale to 16 processors. WindowsNT can scale to
> : > twice that many... Yet "WindowsNT isn't scalable" in the eyes of most
> : > Linux users... silly.
>
>
> All an OEM needs to do, is make minor adjustments to the HAL. WindowsNT,
> according to NTFAQ.com, can support up to 32 processors if the proper
> changes to the HAL are made. If you're referring to "off the shelf"
> WindowsNT, then no, you're limited to 4 processors, unless you're willing
> to deal with an OEM. That fact does not invalidate the fact that
> WindowsNT can support more processors than Linux.
>
> All I'm saying is that the statement "WindowsNT is not scalable" is false.
My point is that, NT which does more than 4 processors is a custom
product that is not a normal configuration. Hell, I'm sure Linux can do
more than 16 with some customization!
>
> : Also, NT clustering is a joke, it is primarily a fail-over cluster.
>
> I have also heard that clustering under WindowsNT v4.0 is less than
> satisfactory. But so what? Progress takes time. If someone really needs
> clustering, they'll use some release of UNIX, until Microsoft releases an
> OS which can handle it properly.
This was my point.
>
> : Linux on the other hand has can do hundreds of computers and a Linux
> : cluster has been rated as one of the fastest computers in the world.
>
> But Linux isn't the only OS with this capability. Most commercial UNIX
> variants can do this sort of thing very well.
No you're right, but, the point was that NT doesn't.
My point is, if Linux is not ready for the Enterprise, I can't honestly
consider NT as ready either. I think Linux is more ready than NT.
Tracy R Reed
>ftp.cdrom.com is using FreeBSD over proprietary Unix. Dejanews is using
>Linux. Yahoo (the last I heard) uses a combination of the two. Do these
>qualify as "halfway decent ISP"s?
etoys.com is entirely Linux and mp3.com uses FreeBSD for it's 8 web servers and
around 50 Linux machines for the rest of the site.
--
Tracy Reed http://www.ultraviolet.org
The sticker on the side of the box said "Supported Platforms: Windows 95,
Windows NT 4.0, or better", so clearly Linux was a supported platform.
Craig Kelley
> : We actually use Apache on two of our NT boxes, instead of IIS (which
> : is also "free"). Admittedly, this was because we already knew how to
> : get perl to work with it. It was nice to simply be able to copy our
> : entire web database system from the AIX box to the NT box and have it
> : all run flawlessly. (that, and the new SMC was more difficult to
> : understand than httpd.conf)
>
> How does Apache stand up on NT, performance-wise?
Fine -- the dual pentiumII/uwide SCSI system is inherently faster than
our old AIX box, so I can't do any real comparisions. It is a patient
analyzing pice for medicade all running under SSL/Apache -- it even
generates graphs on the fly using perl's GD module (something which we
couldn't get to work under IIS -- it would crap out trying to use a
program which spit out image/gif, part of this is IIS/NT's reliance on
filename extensions, so the image tag would have a .PL after it while
the data generated was really a GIF -- Apache doens't do this to us,
it uses the UNIX magic #! to see what to do instead, so our .GIF file
was actually a perl program).
[snip]
Tracy R Reed
>That's a laugh. For an example of a Linux/Apache site, try MP3.com. I'd be
>willing to bet that their daily throughput, as measured in bytes served, is in
>the top 50 sites worldwide (probably top 10, excluding porn). Linux has already
>won... it's just a matter time.
Actually, we peak at 65Mb/s and never go below 30Mb/s. If you figure our
average is 47.5 (just guessing) then we transfer 4104000Mb/day. That's
513000MB/day which is around 513G. My math may be wrong, but you get the idea.
It's a LOT of data and it is growing rapidly. According to
http://slashdot.org/articles/98/11/01/177215.shtml the record set by
ftp.cdrom.com was 759GB. But that's a single FreeBSD box, whereas we are using
many FreeBSD/Linux systems (10 FreeBSD for historical reasons, and another 50
or so Linux systems) because we distribute our content over all of them for
reasons of redundancy and because you can only conveniently fit so much disk in
a single rack mount case.
We currently have 20 Linux boxes dedicated to pushing free mp3's 24/7. There
are many more machines supporting other parts of the site though.
Tracy Reed
System Administrator
tr...@mp3.com
aka
--
Tracy Reed http://www.ultraviolet.org
The sticker on the side of the box said "Supported Platforms: Windows 95,
Windows NT 4.0, or better", so clearly Linux was a supported platform.
I don't speak for mp3.com.
Tracy R Reed
>According to Netcraft's Web Server Survey(http://www.netcraft.com/cgi-bin/Survey/whats),
>mp3.com is running on FreeBSD-Apache, not Linux-Apache. Netcraft
>does get incorrect statistics for some web sites, however.
The static HTML servers are FreeBSD. Try ds1-14 and the many other machines you
may encounter while cruising the site. They are Linux.
Boris
those results certified and put them on www.tpc.org . We'll go from there. It's well known
that lots of so called UNIX people are too dumb to master NT. So they spread all kinds of
lies about it. It's very understandable: an ass will try to blame somebody else for his
own stupidity.
Boris
>Boris, I worked with PC Week on their Linux 2.2 + Samba
>benchmark on a VA Research machjine also. That box had 4
>400 MHz Xeon processors and 2 GB of memory with 5 disks
>in a raid 5 configuration.
>
>The NetBench results for NT serving NT clients were *so
>bad* that PC Week didn't want to publish them. The reason
>it turned out was that NT *will not* use more than 300Mb
>of memory as a file system cache. I was able to tune Linux
>to use 1.8Gb as file system cache.
>
>I love the "days of file server only are mostly gone"
>comment. Isn't that a "well we cannot win in that
>area so lets pretend it doesn't count" comment :-) :-).
>
>What are you going to say when Oracle publishes
>benchmark numbers on Linux. Will it be "days of
>database server are mostly gone.....".
>
>Most NT implementations in my experience are
>used as small file and print servers. If it isn't
>much good for that, what *is* it good for ? Obviously
>not Web serving, Apache holds <50% market share
>there....
>
>So tell me again how "enterprise ready" NT is ? But only
>if you don't have more that 512Mb of Ram maybe ?
>
>BTW: The 300Mb limit is actually *documented* in the
>Windows NT Internals book (2nd edition).
>
>
>Regards,
>
> Jeremy Allison.
bme...@bruce.cs.monash.edu.au
>mlw <ma...@mohawksoft.com> wrote:
>: What released version of NT can scale to more that 4 processors?
>All an OEM needs to do, is make minor adjustments to the HAL. WindowsNT,
>according to NTFAQ.com, can support up to 32 processors if the proper
>changes to the HAL are made. If you're referring to "off the shelf"
>WindowsNT, then no, you're limited to 4 processors, unless you're willing
>to deal with an OEM. That fact does not invalidate the fact that
>WindowsNT can support more processors than Linux.
Does the name "AP1000" ring a bell? All someone had to do was to make minor
adjustments to the linux kernel. The fact that the off-the-funet kernel
will not support more than 8 processors does not invalidate the fact that
linux can support more processors than NT.
Don't you think this is silly?
>All I'm saying is that the statement "WindowsNT is not scalable" is false.
*Grin* Here is a great URL talking about MS's scalability day:
http://www.uniforum.org.nz/newsletters/Jul97.html
Sure, it isn't all that recent, but it's still a great read ;-)
Bernie
--
============================================================================
"It's a magical world, Hobbes ol' buddy...
...let's go exploring"
Calvin's final words, on December 31st, 1995
Mark Hamstra
> Jeremy, don't measure NT which you are ignorant about. You measure your crap (Linux), have
> those results certified and put them on www.tpc.org . We'll go from there. It's well known
> that lots of so called UNIX people are too dumb to master NT. So they spread all kinds of
> lies about it. It's very understandable: an ass will try to blame somebody else for his
> own stupidity.
Heh, heh... nothing like broadcasting your own ignorance, Boris.
The Jeremy Allison that is the target of your ad hominem attack
is one of the key architects of Samba. As a result of his
extensive and long term efforts on that project, he is one of
the world's foremost experts on SMB file sharing and has as
much knowledge of NT's implementation of SMB as anyone outside
of Microsoft (and a good deal more knowledge about such than
most inside Microsoft.)
The unwarranted assertion that Jeremy is ignorant about NT is
truly laughable.
bme...@bruce.cs.monash.edu.au
>Jeremy, don't measure NT which you are ignorant about. You measure
>your crap (Linux), have those results certified and put them on
>www.tpc.org . We'll go from there. It's well known that lots of so
>called UNIX people are too dumb to master NT. So they spread all
>kinds of lies about it. It's very understandable: an ass will try to
>blame somebody else for his own stupidity.
Well, Boris --- I guess _you_ will be ever so glad to provide certified
TPC-D results for Microsoft products. Because by your own logic the lack
of any such results means that nobody would even consider using the products.
Ed
the other way around, but never that one.
Cheers,
ha ha ha ha :):):)
e.
Boris wrote in message <370c4843$0$2...@nntp1.ba.best.com>...
>Jeremy, don't measure NT which you are ignorant about. You measure your
crap (Linux), have
>those results certified and put them on www.tpc.org . We'll go from there.
It's well known
>that lots of so called UNIX people are too dumb to master NT. So they
spread all kinds of
....[arsebiscuits]
Stephen Edwards
: >: What released version of NT can scale to more that 4 processors?
: >All an OEM needs to do, is make minor adjustments to the HAL. WindowsNT,
: >according to NTFAQ.com, can support up to 32 processors if the proper
: >changes to the HAL are made. If you're referring to "off the shelf"
: >WindowsNT, then no, you're limited to 4 processors, unless you're willing
: >to deal with an OEM. That fact does not invalidate the fact that
: >WindowsNT can support more processors than Linux.
: Does the name "AP1000" ring a bell? All someone had to do was to make minor
: adjustments to the linux kernel. The fact that the off-the-funet kernel
: will not support more than 8 processors does not invalidate the fact that
: linux can support more processors than NT.
I must be stuck in the older Linux kernel paradigm. :) I forgot, that
Linux actually has advanced to v2.2, and that things have probably
changed. Originally, I was intending to dispel the myth that WindowsNT is
not scalable. It is scalable, but until Microsoft, or someone else
decides to use WindowsNT to target the high-end, we won't see WindowsNT
running on 16+ processor machines any time soon.
: Don't you think this is silly?
Honestly, yes, a little. :) Though I try to keep an unbiased approach,
many arguments still come down to "my 'puter can beat up your 'puter, so
THPPP!".
: >All I'm saying is that the statement "WindowsNT is not scalable" is false.
: *Grin* Here is a great URL talking about MS's scalability day:
: http://www.uniforum.org.nz/newsletters/Jul97.html
: Sure, it isn't all that recent, but it's still a great read ;-)
*snicker*
"Benchmarketing"... I like it. Personally, I have a hard time believing
benchmarks, simply because I wonder if those who are hosting them have a
hidden agenda of some sort.
This is why I just stick to the philosophy of "use what you like". It
makes life so much more simple.
: Bernie
Bernie Meyer? Didn't you used to frequent the Amiga advocacy group? Are
you to one whom I sniped at for posting "DOS Rules", because I thought I
had posted it before you, only to find that you did indeed post it before
I did?
Stephen Edwards
: Jeremy, don't measure NT which you are ignorant about. You measure your
: crap (Linux), have those results certified and put them on www.tpc.org .
: We'll go from there. It's well known that lots of so called UNIX people
: are too dumb to master NT. So they spread all kinds of lies about it.
: It's very understandable: an ass will try to blame somebody else for his
: own stupidity.
While I agree with you somewhat, Boris, I don't think I would be as
heavy-handed as you in delivery. :)
But you're right, there are many a UNIX user who would rather just abandon
WindowsNT, and conclude that "it's crap", rather than understand what it
is exactly that they are doing wrong. I've made a few BSD users eat their
words in my time. Not many, but a few. :)
I think one barrier between UNIX and NT users is that WindowsNT users
don't understand how _SIMPLE_ UNIX really is. By the same token, most
UNIX users are misled by EXPLORER.EXE, into thinking that WindowsNT
_should_ be simple, when it's really very _COMPLEX_.
Jeremy Allison
>My main problem with benchmarks of any sort are "was each operating system
>configured optimally, and fairly?" I'm not accusing anyone of lying, but
>I do wonder if most of these tests are fair. It's quite clear that many
>UNIX experts are rather ignorant about WindowsNT, as many WindowsNT
>experts are equally ignorant about UNIX.
In a previous job I ported several large UNIX apps to
NT, and was responsible for ensuring that they were
scalable on NT. I am aware at the Win32 level of how to attempt
to get NT to perform well.
Unfortunately, the WinNT SMB server code is closed source,
and there is *ONE* dialog box to tune its performance
(the "optimise as a file server" dialog). Yes, we clicked that :-)
We also set the "don't give preference to foreground
apps" dialog, gave NT 2GB of swap space on a drive
not being used in the test and then we started looking in
the registry for SMB parameters to tune.
Guess what. The thing isn't documented well enough to
let you know how to add more memory for file cache.
If *you* know how to make NT use more than 300MB,
well I'm all ears....
>How are we to be certain that these tests are done fairly, Jeremy?
How can you be certain that *any* test you did not
participate in was done fairly ?
And yet people quote magazine articles.....
I worked with NT people at PC Week, who publish
NT benchmarks for a living. So even if you don't
trust me, you have them to complain at :-).
>I agree that Apache is quite a nice piece of software, but let's be
>realistic here... partly why Apache has taken hold so easily is it's cost.
>Who wants to shell out 1,000's of $ for UNIX Web server software, when
>there is an excellent solution for free?
Errr. Last time I checked IIS is being given
away by Microsoft for free. I don't think
cost is an issue here.
>: BTW: The 300Mb limit is actually *documented* in the
>: Windows NT Internals book (2nd edition).
>Which version of WindowsNT does this book refer to? WindowsNT v4.0 has
>had major changes made to it from v3.5.
The second edition covers NT 4.0. I ask again, what
setting do *you* tune to make NT use more than 300MB for
file cache memory.
Regards,
Jeremy Allison.
Jeremy Allison
>I see. It could be that Microsoft was more concerned with memory
>avaialble to productivity applications than they were memory for other
>tasks.
>Either way, it definitely is a hinderance.
A *hinderence* ? It's a *DISASTER* for file serving.
It forces NT to hit the disk. Hard. That's why
it was so slow. And why does Microsoft have to make
the decision on how much memory you can allocate
to apps or file cache ? What happened to allowing
the sysadmin to choose ? Linux alows me to choose !
Oh yeah, I forgot. NT is "so easy to use" that
NT admins don't have to worry their pretty little
heads about that, Microsoft has made *all* the right
choices for them... :-).
Regards,
Jeremy Allison.
Stephen Edwards
: Stephen Edwards <ja20...@primenet.com> writes:
: apps" dialog, gave NT 2GB of swap space on a drive
Hmmm... well, let me toss this in your lap. Typically, swapspace on a
WindowsNT box should be set up as follows:
1.) One pagefile per fixed disk (_disk_, not partition).
2.) Each pagefile should be set to <physical RAM size> + 12.
Having too much swapspace made available to WindowsNT will actually slow
WindowsNT down a _LOT_. Were there issues at work, that I'm not aware of,
which required such a large amount of swapspace?
: Guess what. The thing isn't documented well enough to
: let you know how to add more memory for file cache.
The registry?... It's documented pretty well, but I will admit that there
is a lot of registry information which gets transferred by word-of-mouth.
I for one have actually learned much about the registry just by "looking"
at it. It does make sense, if you take time to see how it's laid out.
: >How are we to be certain that these tests are done fairly, Jeremy?
: How can you be certain that *any* test you did not
: participate in was done fairly ?
LOL!@# Indeed. :)
: I worked with NT people at PC Week, who publish
: NT benchmarks for a living. So even if you don't
: trust me, you have them to complain at :-).
I'll be the first to admit that WindowsNT has flaws (all OSen do). I just
like to keep things fair, in order to keep FUD, and rumors from
determining "what's good and what's bad".
: >I agree that Apache is quite a nice piece of software, but let's be
: >realistic here... partly why Apache has taken hold so easily is it's cost.
: >Who wants to shell out 1,000's of $ for UNIX Web server software, when
: >there is an excellent solution for free?
: Errr. Last time I checked IIS is being given
: away by Microsoft for free. I don't think
: cost is an issue here.
I think you misunderstand my point. I'm referring to UNIX, in reference
to "why Apache has 50% of the market". Commercial Web servers for UNIX
cost quite a bit of money, last I checked.
: >: BTW: The 300Mb limit is actually *documented* in the
: >: Windows NT Internals book (2nd edition).
: >Which version of WindowsNT does this book refer to? WindowsNT v4.0 has
: >had major changes made to it from v3.5.
: The second edition covers NT 4.0. I ask again, what
: setting do *you* tune to make NT use more than 300MB for
: file cache memory.
NO! I'm sorry, but since you failed in your quest for the sacred
OracleOfTheNTRegistry(tm)(R)(C), you must give pennance by serving
20,000 users with a Timex Sinclair 1000. :-D
Seriously though, I personally don't know how to go about doing that. I
was just suggesting that there might be a way to do so under WindowsNT
v4.0, which perhaps you or I had not been made aware of. There
very well may not be a way... I honestly don't know.
matt
> Boris <XXXb...@movil.comXXX> wrote:
>
> : Jeremy, don't measure NT which you are ignorant about. You measure your
> : crap (Linux), have those results certified and put them on www.tpc.org .
> : We'll go from there. It's well known that lots of so called UNIX people
> : are too dumb to master NT. So they spread all kinds of lies about it.
> : It's very understandable: an ass will try to blame somebody else for his
> : own stupidity.
>
> While I agree with you somewhat, Boris, I don't think I would be as
> heavy-handed as you in delivery. :)
>
> But you're right, there are many a UNIX user who would rather just abandon
> WindowsNT, and conclude that "it's crap", rather than understand what it
> is exactly that they are doing wrong. I've made a few BSD users eat their
> words in my time. Not many, but a few. :)
>
You are right Linux users have not mastered NT neither has even the best NT
admin. We can't. NT is closed source. We can not go in and look at the NT
source code to figure out EXACTLY what happens when you issue a command. It
is not that we are so stupid that we can not understand NT, It is that we
refuse to be stay at the level of stupidity required to be happy with NT.
Jeremy Allison
>Jeremy Allison <jer...@netcom.com> wrote:
>1.) One pagefile per fixed disk (_disk_, not partition).
>2.) Each pagefile should be set to <physical RAM size> + 12.
>Having too much swapspace made available to WindowsNT will actually slow
>WindowsNT down a _LOT_. Were there issues at work, that I'm not aware of,
>which required such a large amount of swapspace?
Thanks for that info. We actually tried it both ways though,
with approx 128Mb swap and also with 2GB swap. Didn't
make a difference.
>The registry?... It's documented pretty well, but I will admit that there
>is a lot of registry information which gets transferred by word-of-mouth.
>I for one have actually learned much about the registry just by "looking"
>at it. It does make sense, if you take time to see how it's laid out.
I know. I discovered how the NT password hashes are stored
in the SAM by "looking" at it. The problem with the registry
is that default values usually aren't stored there at all.
Thus not seeing a key or value doesn't tell you anything
about if a particular key or value actually might exist,
or what potential keys may be possible. That's what I meant
by "undocumented".
>I think you misunderstand my point. I'm referring to UNIX, in reference
>to "why Apache has 50% of the market". Commercial Web servers for UNIX
>cost quite a bit of money, last I checked.
Well Apache has 50% of the *total* Web server
market, not just UNIX only.
>: The second edition covers NT 4.0. I ask again, what
>: setting do *you* tune to make NT use more than 300MB for
>: file cache memory.
>NO! I'm sorry, but since you failed in your quest for the sacred
>OracleOfTheNTRegistry(tm)(R)(C), you must give pennance by serving
>20,000 users with a Timex Sinclair 1000. :-D
If there was a document anywhere outside of Microsoft
documenting it, then I'd like to know about it....
>Seriously though, I personally don't know how to go about doing that. I
>was just suggesting that there might be a way to do so under WindowsNT
>v4.0, which perhaps you or I had not been made aware of. There
>very well may not be a way... I honestly don't know.
I'm coming to the conclusion that this *is* a hard
limit in WinNT4.x. I'd be glad to be proven wrong
though.
Regards,
Jeremy Allison.
Stephen Edwards
: Stephen Edwards <ja20...@primenet.com> writes:
: >Having too much swapspace made available to WindowsNT will actually slow
: >WindowsNT down a _LOT_. Were there issues at work, that I'm not aware of,
: >which required such a large amount of swapspace?
: Thanks for that info. We actually tried it both ways though,
: with approx 128Mb swap and also with 2GB swap. Didn't
: make a difference.
Hmmm... that's odd. Whenever I install WindowsNT on any workstation for
the first time, whether it has 32MB of RAM, or 512MB of RAM, I notice an
immense increase in performance after setting the swapfiles properly.
Those swapfiles were static, were they not? That will also make a huge
difference. :)
Can you tell me what sort of hardware you were using (ie: motherboard,
controllers, display adapters, etc.) WindowsNT is very picky about what
hardware it likes to live in, and the littlest details can make a big
difference in WindowsNT's performance.
: >The registry?... It's documented pretty well, but I will admit that there
: >is a lot of registry information which gets transferred by word-of-mouth.
: >I for one have actually learned much about the registry just by "looking"
: >at it. It does make sense, if you take time to see how it's laid out.
: I know. I discovered how the NT password hashes are stored
: in the SAM by "looking" at it. The problem with the registry
: is that default values usually aren't stored there at all.
Ah. You're referring to the creation of keys/hives, and such. Yes, it's
troublesome. Unless I'm mistaken, I think 3rd party software authors can
have their own keys generated within WindowsNT, with unique parameters,
making the matter even more confusing.
: >: The second edition covers NT 4.0. I ask again, what
: >: setting do *you* tune to make NT use more than 300MB for
: >: file cache memory.
: >NO! I'm sorry, but since you failed in your quest for the sacred
: >OracleOfTheNTRegistry(tm)(R)(C), you must give pennance by serving
: >20,000 users with a Timex Sinclair 1000. :-D
: If there was a document anywhere outside of Microsoft
: documenting it, then I'd like to know about it....
: >Seriously though, I personally don't know how to go about doing that. I
: >was just suggesting that there might be a way to do so under WindowsNT
: >v4.0, which perhaps you or I had not been made aware of. There
: >very well may not be a way... I honestly don't know.
: I'm coming to the conclusion that this *is* a hard
: limit in WinNT4.x. I'd be glad to be proven wrong
: though.
I think I'm going to dig thru some white papers tonight, and see what I
can find... it doesn't make sense to me that Microsoft would make such a
limit... granted, Microsoft has done some silly things for compatability
sakes, but this doesn't make sense to me at all... of course, far be it
from me to suggest that Microsoft embodies perfect logic altogether. :)
jedi
>Boris <XXXb...@movil.comXXX> wrote:
>
>: Jeremy, don't measure NT which you are ignorant about. You measure your
>: crap (Linux), have those results certified and put them on www.tpc.org .
>: We'll go from there. It's well known that lots of so called UNIX people
>: are too dumb to master NT. So they spread all kinds of lies about it.
>: It's very understandable: an ass will try to blame somebody else for his
>: own stupidity.
>
>While I agree with you somewhat, Boris, I don't think I would be as
>heavy-handed as you in delivery. :)
>
>But you're right, there are many a UNIX user who would rather just abandon
>WindowsNT, and conclude that "it's crap", rather than understand what it
>is exactly that they are doing wrong. I've made a few BSD users eat their
But of course. NT is supposed to be the 'easy server OS'. If
it's not easy why bother? I could just run the real VMS and
get remarkably better stability, or even run Unix if I want
to make it easy on myself.
>words in my time. Not many, but a few. :)
>
>I think one barrier between UNIX and NT users is that WindowsNT users
>don't understand how _SIMPLE_ UNIX really is. By the same token, most
>UNIX users are misled by EXPLORER.EXE, into thinking that WindowsNT
>_should_ be simple, when it's really very _COMPLEX_.
If it's no less complex, why bother?
--
"I was not elected to watch my people suffer and die |||
while you discuss this invasion in committe." / | \
In search of sane PPP docs? Try http://penguin.lvcm.com
Jeremy Allison
>Hmmm... that's odd. Whenever I install WindowsNT on any workstation for
>the first time, whether it has 32MB of RAM, or 512MB of RAM, I notice an
>immense increase in performance after setting the swapfiles properly.
>Those swapfiles were static, were they not? That will also make a huge
>difference. :)
I first ran with the default swapfiles as installed. Then
took the advice of the "System" applet on what swap to
set up. Then I tried a "huge" file. All these files were static.
It didn't affect the NetBench performance in any statistically
significant way.
>Can you tell me what sort of hardware you were using (ie: motherboard,
>controllers, display adapters, etc.) WindowsNT is very picky about what
>hardware it likes to live in, and the littlest details can make a big
>difference in WindowsNT's performance.
It was a 4 way 400 MHz Xeon Intel motherboard, using two
dual port 100MB ethernet Intel network cards. The RAID
controller was a Mylex (can't remember the actual copntroller
number). All pretty standard stuff. Linux had drivers for
all of it. I don't even remember what the display card was
as we didn't run Linux with X (it was configured as
a server so we say no need to set up graphics).
>Ah. You're referring to the creation of keys/hives, and such. Yes, it's
>troublesome. Unless I'm mistaken, I think 3rd party software authors can
>have their own keys generated within WindowsNT, with unique parameters,
>making the matter even more confusing.
It isn't the third party apps I care about. It's the
undocumented tuning or configuration parameters
that Microsoft use that I care about. I don't for
a moment believe that all of these are documented.
>I think I'm going to dig thru some white papers tonight, and see what I
>can find... it doesn't make sense to me that Microsoft would make such a
>limit... granted, Microsoft has done some silly things for compatability
>sakes, but this doesn't make sense to me at all... of course, far be it
>from me to suggest that Microsoft embodies perfect logic altogether. :)
Thanks for your help looking for this. I really
appreciate it. Please let me know if you find *anything*
that will change this on NT 4.0.
Thanks,
Jeremy Allison.
Jim Richardson
Stephen Edwards, in the persona of <ja20...@primenet.com>,
brought forth the following words...:
>mlw <ma...@mohawksoft.com> wrote:
>: Stephen Edwards wrote:
>: >
>: > Boris <XXXb...@movil.comXXX> wrote:
>: >
>: > : The latest NT Magazine: April 99.
>: > : Marks claims that Linux 2.2 doesn't scale sufficiently well.
>: >
>: > IIRC, Linux can only scale to 16 processors. WindowsNT can scale to
>: > twice that many... Yet "WindowsNT isn't scalable" in the eyes of most
>: > Linux users... silly.
>
>: What released version of NT can scale to more that 4 processors?
>
>All an OEM needs to do, is make minor adjustments to the HAL. WindowsNT,
>according to NTFAQ.com, can support up to 32 processors if the proper
>changes to the HAL are made. If you're referring to "off the shelf"
>WindowsNT, then no, you're limited to 4 processors, unless you're willing
>to deal with an OEM. That fact does not invalidate the fact that
>WindowsNT can support more processors than Linux.
>
>All I'm saying is that the statement "WindowsNT is not scalable" is false.
are we going to have to go over the difference of "scalable to" vs
"supports" with you again.
Are you claiming that NT *scales* to 32 processors?
>
>: Also, NT clustering is a joke, it is primarily a fail-over cluster.
>
>I have also heard that clustering under WindowsNT v4.0 is less than
>satisfactory. But so what? Progress takes time. If someone really needs
>clustering, they'll use some release of UNIX, until Microsoft releases an
>OS which can handle it properly.
>
>: Linux on the other hand has can do hundreds of computers and a Linux
>: cluster has been rated as one of the fastest computers in the world.
>
>But Linux isn't the only OS with this capability. Most commercial UNIX
>variants can do this sort of thing very well.
>
and NT, cannot.
Jim Richardson
www.eskimo.com/~warlock
All hail Eris
"Linux, because a cpu is a terrible thing to waste."
Stephen Edwards
: On 7 Apr 1999 15:32:19 GMT,
: brought forth the following words...:
: >mlw <ma...@mohawksoft.com> wrote:
: >: Stephen Edwards wrote:
: >: >
: >: > Boris <XXXb...@movil.comXXX> wrote:
: >: >
: >: > : The latest NT Magazine: April 99.
: >: > : Marks claims that Linux 2.2 doesn't scale sufficiently well.
: >: >
: >: > IIRC, Linux can only scale to 16 processors. WindowsNT can scale to
: >: > twice that many... Yet "WindowsNT isn't scalable" in the eyes of most
: >: > Linux users... silly.
: >
: >: What released version of NT can scale to more that 4 processors?
: >
: >All an OEM needs to do, is make minor adjustments to the HAL. WindowsNT,
: >according to NTFAQ.com, can support up to 32 processors if the proper
: >changes to the HAL are made. If you're referring to "off the shelf"
: >WindowsNT, then no, you're limited to 4 processors, unless you're willing
: >to deal with an OEM. That fact does not invalidate the fact that
: >WindowsNT can support more processors than Linux.
: >
: >All I'm saying is that the statement "WindowsNT is not scalable" is false.
: are we going to have to go over the difference of "scalable to" vs
: "supports" with you again.
: Are you claiming that NT *scales* to 32 processors?
Read much? Let me say it again.
"All I'm saying is that the statement "WindowsNT is not scalable" is
false."
Okay, where did I claim WindowsNT can _scale_ to 32 processors?
I said it _____SUPPORTS_____ up to 32 processors.
Akop Pogosian
> The latest NT Magazine: April 99.
> Marks claims that Linux 2.2 doesn't scale sufficiently well.
Of course Linux does not scale well! This is why Avalon and some other Top500
supercomputers happen to run Linux. IBM engineers recently took 17 stock PCs
and built a Beowulf cluster that matched a Cray supercomputer at PovRay
benchmark. And if you are looking for a server (not a cluster of parallel
computers) and need more power than a quad Xeon can give you, you should not
be looking at Intel hardware, Linux or NT anyways.
> This makes sense: OS intended for poor students with 386 cannot be suitable for the
> enterprise.
Yes, only poor people use it. Poor students run it on their 386 computers
(show me one student who still uses a 386, my friend happens to have a dual
PII-350). Poor scientists at universities and NASA build Beowulf clusters with
Linux. Poor IT departments use Linux file and mail servers and don't pay
"per seat" licensing fees. However, I believe the greatest thing about Linux
is not "free beer" like many people think of it, it is more about the "freedom
of speech".
> From my personal experience with several ISPs freeware UNIXes aren't ready for WEB hosting
> either.
Exactly! This is why www.yahoo.com, www.hotmail.com, www.dejanews.com
www/ftp.cdrom.com (note: all of them are killer sites) www.real.com, etc, etc
all run on free unices. Oh, and hotmail belongs to Microsoft. hmmm. And I
thought that MOST small and medium sized ISPs run on free unices anyways.
Last time I read in news about Linux vs NT comparisons in file/web serving
Linux won by a wide margin.. (Was that ZDLabs tests?)
PS: By the way, what is sysinternals.com? Their web site looks like a
"windows only" shop to me. What kind of advice can they give us about any unix
at all? What kind of advice would you _expect_ from them anyways?
David J. Owens
> You are right Linux users have not mastered NT neither has even the best NT
> admin. We can't. NT is closed source. We can not go in and look at the NT
> source code to figure out EXACTLY what happens when you issue a command. It
> is not that we are so stupid that we can not understand NT, It is that we
> refuse to be stay at the level of stupidity required to be happy with NT.
If you must look at the source code to use any operating system or
application software, there is something seriously wrong with the
software's interface or the interface's documentation. You do not
need to see the schematics for your television to change the channel.
You do not have to see the design of the engine to drive your car.
If you are not a developer of the operating system, you should never
be required to see the source code. Period.
Jerry
Nice try with the analogy - but if I chose BBC1 on my TV and it gave me
BBC2 and then a black screen I'd probably take it to be fixed. With a PC
running NT I can't take it to be fixed and I can't find out what its
doing either. Your analogy is therefore very flawed....
I agree that I should never need to see the source code - but I
should'nt have to put up with the crap that Windoze does either - and
being able to see the source code might enable me to fix it - something
which MS seem to be incapable of doing.....
David J. Owens
> David J. Owens wrote:
> > If you must look at the source code to use any operating system or
> > application software, there is something seriously wrong with the
> > software's interface or the interface's documentation. You do not
> > need to see the schematics for your television to change the channel.
> > You do not have to see the design of the engine to drive your car.
> > If you are not a developer of the operating system, you should never
> > be required to see the source code. Period.
>
> Nice try with the analogy - but if I chose BBC1 on my TV and it gave me
> BBC2 and then a black screen I'd probably take it to be fixed. With a PC
> running NT I can't take it to be fixed and I can't find out what its
> doing either. Your analogy is therefore very flawed....
I don't understand what you are saying here. No, you _can_ have
NT fixed. Microsoft issues Service Packs to fix bugs found in
NT's source code. If the problem is with the PC's hardware,
you could send it back to the manufacturer.
There is nothing wrong with this analogy. Television users are not
expected to know anything about how a television works inside.
As a result, television manufacturers must provide a well designed
and documented interface for using the television. In my opinion,
Linux's interface is poorly designed and documented. By interface
I mean the total suite of programs available for installing, configuring,
and maintaining the operating system. Linux users may say that
Linux's tools are well documented due to the large amount of documentation
available in man pages, FAQs, and How-Tos. However, I have
found that these sources of information to are either written for
someone who already knows the utilities well(man pages) or are
just too difficult to understand easily(FAQs, How-Tos). In any
case, if Linux's interface was as well designed as that of NT,
Linux would not need this extreme quantity of documentation.
> I agree that I should never need to see the source code - but I
> should'nt have to put up with the crap that Windoze does either - and
> being able to see the source code might enable me to fix it - something
> which MS seem to be incapable of doing.....
In my experience with NT and Linux, you can expect to have
far more problems with Linux than NT. Linux is like one of
those run-down houses for sale called a "Handyman's Dream"
by the real estate company. You can make it into something
decent, but you may have to do the work yourself(by using the
Linux source code).
Jerry
>
> Jerry wrote:
>
> > David J. Owens wrote:
> > > If you must look at the source code to use any operating system or
> > > application software, there is something seriously wrong with the
> > > software's interface or the interface's documentation. You do not
> > > need to see the schematics for your television to change the channel.
> > > You do not have to see the design of the engine to drive your car.
> > > If you are not a developer of the operating system, you should never
> > > be required to see the source code. Period.
> >
> > Nice try with the analogy - but if I chose BBC1 on my TV and it gave me
> > BBC2 and then a black screen I'd probably take it to be fixed. With a PC
> > running NT I can't take it to be fixed and I can't find out what its
> > doing either. Your analogy is therefore very flawed....
>
> I don't understand what you are saying here. No, you _can_ have
> NT fixed. Microsoft issues Service Packs to fix bugs found in
> NT's source code. If the problem is with the PC's hardware,
> you could send it back to the manufacturer.
Oh yeah - how come NT still has so many bugs then ? How come Service
Pack 4 - required for Y2K compliance - seems to introduce more bugs than
it fixes, how come W98 still will not run for more than 49.7 days - and
so on - how long will it take MS to fix these (they has to be forced to
produce SP4 by coprorate industry - how much chance have I got at
getting a bug fixed ?
>
> There is nothing wrong with this analogy. Television users are not
> expected to know anything about how a television works inside.
> As a result, television manufacturers must provide a well designed
> and documented interface for using the television.
Yes they do - and MS don't.....The analogy is crap - if my TV "crashed"
as often as my PC it would spend more time back at the repair shop than
in my lounge. I can also repair a TV - I have'nt a hope in hell of
fixing Windoze... If I sent my PC back everytime I found something not
working as documented then I would'nt have a PC - it would arrive at my
house and be despatched back to be fixed within minutes....if running
any MS OS that is.....
> In my opinion,
> Linux's interface is poorly designed and documented.
Pretty unknowledgable opinion then....
> By interface
> I mean the total suite of programs available for installing, configuring,
> and maintaining the operating system.
Yeah great - much of which is badly/wrongly documented or does'nt work
as the documentation says.
> Linux users may say that
> Linux's tools are well documented due to the large amount of documentation
> available in man pages, FAQs, and How-Tos. However, I have
> found that these sources of information to are either written for
> someone who already knows the utilities well(man pages) or are
> just too difficult to understand easily(FAQs, How-Tos).
Strange - that contradicts other NT to Linux converts I know who claim
that the Linux documentation is MUCH more comprehensive and CORRECT than
that supplied by MS...
> In any
> case, if Linux's interface was as well designed as that of NT,
> Linux would not need this extreme quantity of documentation.
Rubbish - the Windoze interface is about as bad as a GUI can get and the
documentation is shit - how come users have so many problems with
Wondoze ? It does'nt make NT better - the fact is that you CAN'T GET the
documentation for NT regardless os what its GUI is like (crap...).
>
> > I agree that I should never need to see the source code - but I
> > should'nt have to put up with the crap that Windoze does either - and
> > being able to see the source code might enable me to fix it - something
> > which MS seem to be incapable of doing.....
>
> In my experience with NT and Linux, you can expect to have
> far more problems with Linux than NT. Linux is like one of
> those run-down houses for sale called a "Handyman's Dream"
> by the real estate company. You can make it into something
> decent, but you may have to do the work yourself(by using the
> Linux source code).
Well - your experience contradicts my own and virtually everybodies in
this group AND the corporate world at the moment. You are virtually on
your own here....... Seems your experience is'nt worth a shit...
Do you work for MS ? You are either VERY ignorant of their products or
payed to "overlook" their failings....
jedi
>Jerry wrote:
>
>> David J. Owens wrote:
>> > If you must look at the source code to use any operating system or
>> > application software, there is something seriously wrong with the
>> > software's interface or the interface's documentation. You do not
>> > need to see the schematics for your television to change the channel.
>> > You do not have to see the design of the engine to drive your car.
>> > If you are not a developer of the operating system, you should never
>> > be required to see the source code. Period.
>>
>> Nice try with the analogy - but if I chose BBC1 on my TV and it gave me
>> BBC2 and then a black screen I'd probably take it to be fixed. With a PC
>> running NT I can't take it to be fixed and I can't find out what its
>> doing either. Your analogy is therefore very flawed....
>
>I don't understand what you are saying here. No, you _can_ have
>NT fixed. Microsoft issues Service Packs to fix bugs found in
>NT's source code. If the problem is with the PC's hardware,
>you could send it back to the manufacturer.
>
>expected to know anything about how a television works inside.
>As a result, television manufacturers must provide a well designed
No, they have to provide a device that actually works
reliably. Microsoft doesn't do this and never gets held
accountable for it.
>Linux's interface is poorly designed and documented. By interface
>I mean the total suite of programs available for installing, configuring,
>and maintaining the operating system. Linux users may say that
Actually, it's quite well documented. Being primarily
a BSD from that point of view is quite an advantage.
Plus, it actually comes with some useful hardcopy
documentation in it's sold forms. Windows doesn't
anymore.
>Linux's tools are well documented due to the large amount of documentation
>available in man pages, FAQs, and How-Tos. However, I have
>found that these sources of information to are either written for
>someone who already knows the utilities well(man pages) or are
Poppycock. manpages are quite adequate for the clueless.
They just require a little balls. Howto's are the same
for the most part with the exception of the badly
written example.
>just too difficult to understand easily(FAQs, How-Tos). In any
>case, if Linux's interface was as well designed as that of NT,
>Linux would not need this extreme quantity of documentation.
For the most part it doesn't. For the vast majority
of pedestrian user problems: use the damn applet is
quite a useful directive.
It's not like Redhat doesn't litter it's default
Administrator desktop with it's control panel.
>
>> I agree that I should never need to see the source code - but I
>> should'nt have to put up with the crap that Windoze does either - and
>> being able to see the source code might enable me to fix it - something
>> which MS seem to be incapable of doing.....
>
>In my experience with NT and Linux, you can expect to have
>far more problems with Linux than NT. Linux is like one of
Like what precisely?
>those run-down houses for sale called a "Handyman's Dream"
>by the real estate company. You can make it into something
>decent, but you may have to do the work yourself(by using the
>Linux source code).
This just a flat out lie. The only sourcecode I've ever
had to touch was a faulty driver which you can't get
away from under NT (SBLive).
The rest is manipulating menus, maybe running a command
line or 2 or editing some files if your chose apps
require it (some do, some don't).
Your information is about 2-4 years out of date.
Stephen Edwards
: >
: > > You are right Linux users have not mastered NT neither has even the best NT
: > > admin. We can't. NT is closed source. We can not go in and look at the NT
: > > source code to figure out EXACTLY what happens when you issue a command. It
: > > is not that we are so stupid that we can not understand NT, It is that we
: > > refuse to be stay at the level of stupidity required to be happy with NT.
: > If you must look at the source code to use any operating system or
: > application software, there is something seriously wrong with the
: > software's interface or the interface's documentation. You do not
: > need to see the schematics for your television to change the channel.
: > You do not have to see the design of the engine to drive your car.
: > If you are not a developer of the operating system, you should never
: > be required to see the source code. Period.
: Nice try with the analogy - but if I chose BBC1 on my TV and it gave me
: BBC2 and then a black screen I'd probably take it to be fixed. With a PC
: running NT I can't take it to be fixed and I can't find out what its
: doing either. Your analogy is therefore very flawed....
You say his analogy is off, but you're logic is very screwed up. What do
you mean "can't take it in to be fixed"?... There are tons of computer
repair services, available on call, barring technical support from an OEM.
Stephen Edwards
: >
: > > David J. Owens wrote:
: > > > If you must look at the source code to use any operating system or
: > > > application software, there is something seriously wrong with the
: > > > software's interface or the interface's documentation. You do not
: > > > need to see the schematics for your television to change the channel.
: > > > You do not have to see the design of the engine to drive your car.
: > > > If you are not a developer of the operating system, you should never
: > > > be required to see the source code. Period.
: > >
: > > Nice try with the analogy - but if I chose BBC1 on my TV and it gave me
: > > BBC2 and then a black screen I'd probably take it to be fixed. With a PC
: > > running NT I can't take it to be fixed and I can't find out what its
: > > doing either. Your analogy is therefore very flawed....
: > I don't understand what you are saying here. No, you _can_ have
: > NT fixed. Microsoft issues Service Packs to fix bugs found in
: > NT's source code. If the problem is with the PC's hardware,
: > you could send it back to the manufacturer.
: Oh yeah - how come NT still has so many bugs then ? How come Service
Which bugs would those be? Care to provide a list of these bugs you've
somehow been able to dig out of the NT source tree? You must be special,
because I for one have NEVER seen one single line of source for WindowsNT.
: Do you work for MS ? You are either VERY ignorant of their products or
: payed to "overlook" their failings....
So how are we to beleive that you are not a moaning ignoramous? You've
made several interesting claims about WindowsNT, yet you have failed to
back any of it up with fact.
Timothy Kelley
wrote:
>Ha ha. Excellent. Unix users not being able to master NT. I've heard of
>the other way around, but never that one.
>
>Cheers,
>ha ha ha ha :):):)
Actually, there's a bit of truth to it. Unix - all of it - is
*understandable*. Every last bit of it, and you don't have to be a
genius to understand it either (just curious).
NT (superficially) seems understandable, but I think the number of
people who really understand it (registry and all) are very few. It
is actually much more complicated than unix (and that isn't a good
thing). Moreover, anyone who bothers to learn all this is probably
wasting their time since Micros~1 will just change everything again
anyway ...
Stephen Edwards
: On Thu, 8 Apr 1999 16:33:31 -0700, "Ed" <e...@no-spam-please-iiu.ie>
: wrote:
: >Ha ha. Excellent. Unix users not being able to master NT. I've heard of
: >the other way around, but never that one.
: Actually, there's a bit of truth to it. Unix - all of it - is
: *understandable*. Every last bit of it, and you don't have to be a
: genius to understand it either (just curious).
You don't have to be a genius to understand WindowsNT either. You just
have to be willing to spend a little more time learning
TheMicrosoftWay(tm). They both make perfect sense to me, and if __I__ can
understand both WindowsNT and UNIX, then __ANYONE__ can.
In short, WindowsNT is a different "way of thinking" than UNIX is.
: NT (superficially) seems understandable, but I think the number of
: people who really understand it (registry and all) are very few. It
: is actually much more complicated than unix (and that isn't a good
: thing). Moreover, anyone who bothers to learn all this is probably
: wasting their time since Micros~1 will just change everything again
: anyway ... ^^^^^^^^
*snicker* :) You actually made me double-check my VT setting for my
terminal.
WindowsNT has changed quite a bit from v3.1 on, but not in any way that's
so difficult to understand. I mean, let's be realistic here... anyone who
is experienced in administering OSen can learn a new OS in about a week.
jedi
>Jerry <jerrys121DEL...@hotmail.com> wrote:
>: David J. Owens wrote:
>: >
>: >
>: > > You are right Linux users have not mastered NT neither has even the best NT
>: > > admin. We can't. NT is closed source. We can not go in and look at the NT
>: > > source code to figure out EXACTLY what happens when you issue a command. It
>: > > is not that we are so stupid that we can not understand NT, It is that we
>: > > refuse to be stay at the level of stupidity required to be happy with NT.
>: >
>: > application software, there is something seriously wrong with the
>: > software's interface or the interface's documentation. You do not
>: > need to see the schematics for your television to change the channel.
>: > You do not have to see the design of the engine to drive your car.
>: > If you are not a developer of the operating system, you should never
>: > be required to see the source code. Period.
>
>: Nice try with the analogy - but if I chose BBC1 on my TV and it gave me
>: BBC2 and then a black screen I'd probably take it to be fixed. With a PC
>: running NT I can't take it to be fixed and I can't find out what its
>: doing either. Your analogy is therefore very flawed....
>
>you mean "can't take it in to be fixed"?... There are tons of computer
>repair services, available on call, barring technical support from an OEM.
...not to the degree he's speaking of.
There are no such services because the information required
to implement them runs in the millions and is otherwise
unatainable.
Stephen Edwards
: On 9 Apr 1999 20:20:14 GMT, Stephen Edwards <ja20...@primenet.com> wrote:
: >Jerry <jerrys121DEL...@hotmail.com> wrote:
: >
: >: BBC2 and then a black screen I'd probably take it to be fixed. With a PC
: >: running NT I can't take it to be fixed and I can't find out what its
: >: doing either. Your analogy is therefore very flawed....
: >
: >You say his analogy is off, but you're logic is very screwed up. What do
: >you mean "can't take it in to be fixed"?... There are tons of computer
: >repair services, available on call, barring technical support from an OEM.
: ...not to the degree he's speaking of.
Well, if you want to be able to depend upon any two-bit data doctor in the
corner shopping center, then neither WindowsNT, nor UNIX apply.
But what he said was nonsense.
: There are no such services because the information required
: to implement them runs in the millions and is otherwise
: unatainable.
If this is so, then how is it that WindowsNT OEMs are able to provide
technical support for their products? Do you think for one minute that
COMPAQ, DELL, or HP do not have a plethora of information about WindowsNT
at their fingertips that you and I will probably never see? If this was
the case, then OEMs wouldn't exist in the first place.
An OEM version of Microsoft software is that which is completely
custom-tailored by an OEM, to suit their make and model of machines. How
do you think OEMs can exist without knowing how to go about customizing
the OS for themselves?
By the way, an OEM does not just refer to large corporations who sell big
iron... there are many OEMs that sell to the end consumer as well.
jedi
>jedi <je...@dementia.mishnet> wrote:
>: On 9 Apr 1999 20:20:14 GMT, Stephen Edwards <ja20...@primenet.com> wrote:
>: >Jerry <jerrys121DEL...@hotmail.com> wrote:
>: >
>: >: Nice try with the analogy - but if I chose BBC1 on my TV and it gave me
>: >: BBC2 and then a black screen I'd probably take it to be fixed. With a PC
>: >: running NT I can't take it to be fixed and I can't find out what its
>: >: doing either. Your analogy is therefore very flawed....
>: >
>: >You say his analogy is off, but you're logic is very screwed up. What do
>: >you mean "can't take it in to be fixed"?... There are tons of computer
>: >repair services, available on call, barring technical support from an OEM.
>
>: ...not to the degree he's speaking of.
>
>Well, if you want to be able to depend upon any two-bit data doctor in the
>corner shopping center, then neither WindowsNT, nor UNIX apply.
Yes it does. There is Linux, there are the free BSDs,
there are the more widely source licenced commercial
unixen as well as BSDi which is cheaper than NT is
for the enterprise despite coming with source.
>
>But what he said was nonsense.
>
>: There are no such services because the information required
>: to implement them runs in the millions and is otherwise
>: unatainable.
>
>If this is so, then how is it that WindowsNT OEMs are able to provide
>technical support for their products? Do you think for one minute that
>COMPAQ, DELL, or HP do not have a plethora of information about WindowsNT
That's tech support versus actually fixing problems. That's
worlds apart from the TV repairman unless your concept of
such a creature is someone who tells you over the phoen to
slap the TV once or twice.
>at their fingertips that you and I will probably never see? If this was
>the case, then OEMs wouldn't exist in the first place.
OEMs exist to sell cheap solutions that require low overhead.
For serious support or actual repairs, their abilities are
somewhat limited.
When's the last time a Dell fixed something like an addressable
memory limitation (VA Research).
>
>An OEM version of Microsoft software is that which is completely
>custom-tailored by an OEM, to suit their make and model of machines. How
>do you think OEMs can exist without knowing how to go about customizing
>the OS for themselves?
They don't they just resell a canned solution for the most part.
>
>By the way, an OEM does not just refer to large corporations who sell big
>iron... there are many OEMs that sell to the end consumer as well.
Those are the most worthless of all.
Rob Hughes
boris...@pleasemovil.com says...
> :)The latest NT Magazine: April 99.
> :)Marks claims that Linux 2.2 doesn't scale sufficiently well.
> :)This makes sense: OS intended for poor students with 386 cannot be suitable for the
> :)enterprise.
> :)From my personal experience with several ISPs freeware UNIXes aren't ready for WEB hosting
> :)either.
> :)
> :)Boris
> :)
If this isn't a troll, there never was one.
-----------== Posted via Newsfeeds.Com, Uncensored Usenet News ==----------
http://www.newsfeeds.com/ The Largest Usenet Servers in the World!
---------== Over 72,000 Groups, Plus Dedicated Binaries Servers ==--------
Grant Fischer
>Read much? Let me say it again.
>
>"All I'm saying is that the statement "WindowsNT is not scalable" is
> false."
>
>Okay, where did I claim WindowsNT can _scale_ to 32 processors?
Read the quotes yourself, especially the part from
message <7eeqsc$257$1...@nnrp02.primenet.com>
Am I reading the quotes wrong, or what?
--
Grant Fischer (gfischer at the domain hub.org)
Jim Richardson
>Read much? Let me say it again.
>
>"All I'm saying is that the statement "WindowsNT is not scalable" is
> false."
>
>Okay, where did I claim WindowsNT can _scale_ to 32 processors?
>
look above, you said:
>: >: > IIRC, Linux can only scale to 16 processors. WindowsNT can scale to
>: >: > twice that many... Yet "WindowsNT isn't scalable" in the eyes of most
>: >: > Linux users... silly.
So are you claiming that NT can scale to 32 processors?
(you will note I hope, that 16*2 is 32...)
>I said it _____SUPPORTS_____ up to 32 processors.
yes, in response to a question of whether NT could _____SCALE____
to more than 4 processors. So you either
a) responded to the question with a non-sequitor or
b) tried to imply that NT could scale beyond 4 processors without actually
saying so...
So I'll ask the question again,
"What released version of NT can scale to more than 4 processors?"
note that I said scale, not support.
I am not trying to flame, or deride, just looking for the answer to
the question
--
Christopher Smith
>
> "What released version of NT can scale to more than 4 processors?"
>
> note that I said scale, not support.
NT Enterprise, supposedly. Of course:
a) there aren't that many *applications* that are going to scale to that
many CPUs.
b) if you want/need >4 (or even >2) CPUs, you really shouldn't be trying to
do it with x86 anyway, be it Linux, NT, Solaris or whatever.
Volker Dittmar
>
> NT is optimized as application server (as opposed to file server). Days of file servers
> are mostly gone.
Not true. In fact, most servers that are NOT serving the internet are file servers.
Where do YOU store your corporate documents???
> Plus ZD test machines in NT vs. Linux results you describe had only 64MB of RAM; this is
> clearly insufficient amount of RAM for production NT server. To make my point clearer: DOS
> beats NT with <16 MB of RAM (because NT cannot even run with less than 16 MB). While there
> are dozens of NT benchmark results on www.tpc.org and www.spec.org there isn't a single
> mentioning of Linux on those sites. Which just proves the thing: nobody uses this thing
> (Linux) as enterprise server platform as opposed to NT (which is used very widely). Mark
> gave theoretical explanation of the fact that Linux isn't usable in enterprise server
> role.
WRONG. SAP just bought parts of RedHat, because THEY already deploy mission critical
enterprise servers running Linux (read their announcement on the CBit).
SAP claims that Linux performs better than NT on the same hardware. Now everybody
who knows SAP software will acknowledge that there is no SAP server with less than
256MB RAM (most have much more RAM). SAP claims that Linux has a better performance
for their apps and is far more stable.
And, PC magazine has an article about the performance of Samba/Linux on machines
with 1GB RAM. They did not post the benchmark figures for NT because NT performed
so poorly that they thought they did soemthing wrong (they didn't manage to let
NT use more than 300MB for their file cache - it is still a mystery how to do this,
even M$ does not know, AFAIK).
Just because there is no TPC-C benchmark makes you state that Linux isn't ready
for this.
There is no TPC benchmark for Linux because nobody needs to adverties it, and
so no one is willing to pay money for testing it. That's just it. But you keep
on saying that this "proves" that Linux won't beat NT.
Ciao,
Volker Dittmar
Anthony Ord
<boris...@pleasemovil.com> wrote:
>NT is optimized as application server (as opposed to file server).
So it's optimised to run applications on? Like a mainframe?
Oh, this is news. What applications run on NT server?
>Days of file servers
>are mostly gone.
Naah. It will always be useful for sharing data. Mail is
good, but stupid within the same company/LAN.
>Plus ZD test machines in NT vs. Linux results you describe had only 64MB of RAM; this is
>clearly insufficient amount of RAM for production NT server.
How much is sufficient for NT to show its superiority?
>To make my point clearer: DOS
>beats NT with <16 MB of RAM (because NT cannot even run with less than 16 MB).
NT 3.51 can. And you can serve more than 10 clients with
NTWS without violating you EULA.
>While there
>are dozens of NT benchmark results on www.tpc.org and www.spec.org there isn't a single
>mentioning of Linux on those sites. Which just proves the thing: nobody uses this thing
>(Linux) as enterprise server platform as opposed to NT (which is used very widely). Mark
>gave theoretical explanation of the fact that Linux isn't usable in enterprise server
>role.
>Boris
Regards
Anthony
<snip>
--
-----------------------------------------
| And when our worlds |
| They fall apart |
| When the walls come tumbling in |
| Though we may deserve it |
| It will be worth it - Depeche Mode |
-----------------------------------------
Anthony Ord
<boris...@pleasemovil.com> wrote:
>Jeremy, don't measure NT which you are ignorant about. You measure your crap (Linux), have
>those results certified and put them on www.tpc.org . We'll go from there. It's well known
>that lots of so called UNIX people are too dumb to master NT. So they spread all kinds of
>lies about it. It's very understandable: an ass will try to blame somebody else for his
>own stupidity.
So how do you get NT to use more than 300MB of memory as a
disk cache, given 2Gb of memory? It's documented, so it must
be changeable.
Is it something obvious?
>Boris
Anthony Ord
>Jim Richardson <war...@eskimo.com> wrote:
>: On 7 Apr 1999 15:32:19 GMT,
>: Stephen Edwards, in the persona of <ja20...@primenet.com>,
>: brought forth the following words...:
>
>: >mlw <ma...@mohawksoft.com> wrote:
>: >: Stephen Edwards wrote:
>: >: >
>: >: > Boris <XXXb...@movil.comXXX> wrote:
>: >: >
>: >: > : The latest NT Magazine: April 99.
>: >: > : Marks claims that Linux 2.2 doesn't scale sufficiently well.
>: >: >
>: >: > IIRC, Linux can only scale to 16 processors. WindowsNT can scale to
>: >: > twice that many... Yet "WindowsNT isn't scalable" in the eyes of most
***************************
>: >: > Linux users... silly.
>: >
>: >: What released version of NT can scale to more that 4 processors?
>: >
>: >All an OEM needs to do, is make minor adjustments to the HAL. WindowsNT,
>: >according to NTFAQ.com, can support up to 32 processors if the proper
>: >changes to the HAL are made. If you're referring to "off the shelf"
>: >WindowsNT, then no, you're limited to 4 processors, unless you're willing
>: >to deal with an OEM. That fact does not invalidate the fact that
>: >WindowsNT can support more processors than Linux.
>: >
>: >All I'm saying is that the statement "WindowsNT is not scalable" is false.
>
>: are we going to have to go over the difference of "scalable to" vs
>: "supports" with you again.
>: Are you claiming that NT *scales* to 32 processors?
>
>Read much? Let me say it again.
>
>"All I'm saying is that the statement "WindowsNT is not scalable" is
> false."
>
>Okay, where did I claim WindowsNT can _scale_ to 32 processors?
>
>I said it _____SUPPORTS_____ up to 32 processors.
You said it where the asterisks are.
Twice as many as 16 is 32. Or at least where I come from.
>[] "No footnote for you!" -- Footnote Nazi
Regards
Anthony
Todd Bandrowsky
>expected to know anything about how a television works inside.
>As a result, television manufacturers must provide a well designed
>and documented interface for using the television.
So how does that explain the clocks on VCRS?
> In any
>case, if Linux's interface was as well designed as that of NT,
>Linux would not need this extreme quantity of documentation.
Rant 0 - Windows is not necessarily easier to use or learn than Linux
----------------------------------------------------------------------------
-------------
I know a lot of end users that prefer the command line to the gui. Which is
easier to say on a phone call, or explain:
a. "ok, now I want you to type ls, and press the enter key."
or
b. "now, I want you to click on the task bar. that's the little bar on
the bottom of the screen. oh, you don't have one on the bottom of the
screen? Well, check to see if it is on the top or on the sides of the
screen. It has a little clock on it, unless you checked it off. No mam, I
don't know what color it is."
"Ok, you found the taskbar. great, now move the mouse over it, and click on
it with the right mouse button. You should get a little menu that. Oh a
menu, that's the list of stuff. One of those things will say "minimize all
windows". Click on that.
"No. no. you didn't lose all of your stuff. It's just hidden away so we can
get to the next step."
"Ok, now click on my computer, and you.."
Digression 0 - GUI Pet Peeves
---------------------------------------
I could go on like this all day, but say that any GUI is somehow inherently
better because they do not have a descriptive mechanism that works worth a
shit.
My two pet peeves:
a. There is nothing on screen that says what different objects are.
These are obvious: what in Windows says what a window is. where does the
little task bar system tray whatever the system tray is.
b. iconography tries to show an abstraction, and should show what the
piece of software actually looks like. If stores sold goods the way that
icons are used to represent tools and programs, then we would have people
buying condoms when they meant to buy soap, buying a car when they meant to
buy toilet paper. Generally, when you see a box, it has a picture of the
thing that you are going to get when you open it. Yet, we brilliant
software people have turned such obvious logic on its head with the likes of
the radar dish that somehow means "view channels", with the goofy looking
assortment of e's and globes and i's that mean the browser. None of it
makes any sense. It took me quite a bit of imagination to understand how
the peer web services icon of handing a piece of paper to the earth somehow
meant world wide web.
that means that every application should look as different as possible, not
the same. It is certainly desirable to act in the same sorts of ways, where
appropriate, but, by and large, everything should look totally different.
People use vision to distinguish things, and what happens when everything
looks the same?
I would strenously argue that it might turn out that the ugly medley that
characterizes X software, so often criticized because it is not consistent,
will actually turn out to be more user friendly for this reason alone....
If I were a Linux advocate, I would argue with a straight face that the X
systems differing and inconsistent approach makes software under Linux
easier to learn because it is different.
That the world is so screwed up is proof that advocates of consistency have
been too successful.
Rant 1, TV versus Windows
-----------------------------------
Well, the problem with your analogy is that nothing stops you from finding
out how the TV or car works. You are perfectly allowed to take it apart, as
are competitors within the general manufacturing industry. Under present
licensing arrangements, you cannot do that with most commercial software.
What is happening in the software community is apalling. We've created a
system where we buy products, but cannot own them, where these products that
are made are not even required to actually work. Users of commercial
software have basically traded away all of their rights.
I know I have some good feelings about Microsoft, and other closed source
vendors, and, I appreciate that you may have them too. But feelings should
not be perceived to be the legal reality, which is stated on the license of
every microsoft product you ever buy:
a. you are not guaranteed that the product will actually work.
b. you are not allowed to take it apart, or modify.
Thus, Microsoft, according to its licenses, is perfectly entitled to sell
you a product that does not work and that you cannot even attempt to fix.
Last time I checked, if I bought a car that didn't work, I can get it fixed.
If I buy a TV that doesn't work, I can return it.
You could argue as much as you want about "the advantages of a supported
product" that goes with the close sourced license system, but, in reality
all you are doing is spending money to make yourself feel good and Bill
Gates, Larry Ellison and that slug over at Sun more rich.
Legally, when you purchase a license, the manufacturer is not compelled to
give you anything.
The only way that a closed source model could ever prevail over the open
source system is for EULA's to give you more.
>
>> I agree that I should never need to see the source code - but I
>> should'nt have to put up with the crap that Windoze does either - and
>> being able to see the source code might enable me to fix it - something
>> which MS seem to be incapable of doing.....
>
>In my experience with NT and Linux, you can expect to have
>far more problems with Linux than NT. Linux is like one of
Todd Bandrowsky
or
My two pet peeves:
b. Iconography tries to show an abstraction, and should show what the
piece of software actually looks like.
In the real world, when you see a box, it has a picture of the thing that
you are going to get when you open it. Yet, we brilliant software people
have turned such obvious logic on its head with the likes of the radar dish
that somehow means "view channels", and have the modern arrangement of the
goofy looking assortment of e's and globes and i's that mean the browser.
None of it makes any sense. It took me quite a bit of imagination to
understand how the peer web services icon of handing a piece of paper to the
earth somehow meant world wide web.
If stores sold goods the way that icons are used to represent tools and
programs, then we would have people buying condoms when they meant to buy
soap, buying a car when they meant to buy toilet paper. Befuddled consumers
would walk around starving but with plenty of bailing wire, would walk
around with two shirts but no shoes.
Were computer programmers designers of packages, industrial society would
grind to a tragic halt as consumers continually purchased the wrong
products.
There are few abstractions that can be adequately visualized - even a disk
does not look like a disk anymore. Instead, the key is to make the
application or tool look different, and then make the icon for that tool
look like the thing they are going to use. Gosh, one should change the
colors of everything, if possible.
People use vision to distinguish things, and what happens when everything
looks the same?
If I were a Linux advocate, I would argue with a straight face that the X
systems differing and inconsistent approach makes software under Linux
easier to learn because it is different.
Rant 1, TV versus Windows
Boris
produce, SAMBA is another).
Boris
Mark Hamstra wrote in message ...
>"Boris" <boris...@pleasemovil.com> writes:
>
>> Jeremy, don't measure NT which you are ignorant about. You measure your crap (Linux),
have
>> those results certified and put them on www.tpc.org . We'll go from there. It's well
known
>> that lots of so called UNIX people are too dumb to master NT. So they spread all kinds
of
>> lies about it. It's very understandable: an ass will try to blame somebody else for his
>> own stupidity.
>
>
>Heh, heh... nothing like broadcasting your own ignorance, Boris.
>The Jeremy Allison that is the target of your ad hominem attack
>is one of the key architects of Samba. As a result of his
>extensive and long term efforts on that project, he is one of
>the world's foremost experts on SMB file sharing and has as
>much knowledge of NT's implementation of SMB as anyone outside
>of Microsoft (and a good deal more knowledge about such than
>most inside Microsoft.)
>
>The unwarranted assertion that Jeremy is ignorant about NT is
>truly laughable.
>
>
>> Boris
>> >Boris, I worked with PC Week on their Linux 2.2 + Samba
>> >benchmark on a VA Research machjine also. That box had 4
>> >400 MHz Xeon processors and 2 GB of memory with 5 disks
>> >in a raid 5 configuration.
>> >
>> >The NetBench results for NT serving NT clients were *so
>> >bad* that PC Week didn't want to publish them. The reason
>> >it turned out was that NT *will not* use more than 300Mb
>> >of memory as a file system cache. I was able to tune Linux
>> >to use 1.8Gb as file system cache.
>> >
>> >I love the "days of file server only are mostly gone"
>> >comment. Isn't that a "well we cannot win in that
>> >area so lets pretend it doesn't count" comment :-) :-).
>> >
>> >What are you going to say when Oracle publishes
>> >benchmark numbers on Linux. Will it be "days of
>> >database server are mostly gone.....".
>> >
>> >Most NT implementations in my experience are
>> >used as small file and print servers. If it isn't
>> >much good for that, what *is* it good for ? Obviously
>> >not Web serving, Apache holds <50% market share
>> >there....
>> >
>> >So tell me again how "enterprise ready" NT is ? But only
>> >if you don't have more that 512Mb of Ram maybe ?
>> >
>> >BTW: The 300Mb limit is actually *documented* in the
>> >Windows NT Internals book (2nd edition).
>> >
>> >
>> >Regards,
>> >
>> > Jeremy Allison.
>
Mark Hamstra
> LINUX crowd is full of crap-producing, self-respected "gurus" (LINUX is just one crap they
> produce, SAMBA is another).
It certainly is odd then how that "crap" Samba running on
that "crap" OS, Linux, soundly outperforms NT's SMB filesharing.
It is olso very odd that the "self-respected" Jeremy Allison
took no public affront to your unwarranted indictment of his
NT knowledge, preferring instead a very low-key approach that
lets the performance numbers speak for themselves.
Or perhaps it is the case that neither Samba nor Linux are
crap, and that Jeremy has sufficient humilty not to be
obsessed with the value of his own opinions and has earned
the respect of myself and many knowledgeable system admin-
istrators.
Mark
Martin Ozolins
Jeremy Allison wrote in message ...
>Stephen Edwards <ja20...@primenet.com> writes:
>
>>Jeremy Allison <jer...@netcom.com> wrote:
>
>
>>1.) One pagefile per fixed disk (_disk_, not partition).
>
>>2.) Each pagefile should be set to <physical RAM size> + 12.
>
>>Having too much swapspace made available to WindowsNT will actually slow
>>WindowsNT down a _LOT_. Were there issues at work, that I'm not aware of,
>>which required such a large amount of swapspace?
>
>Thanks for that info. We actually tried it both ways though,
>with approx 128Mb swap and also with 2GB swap. Didn't
>make a difference.
>
>>The registry?... It's documented pretty well, but I will admit that there
>>is a lot of registry information which gets transferred by word-of-mouth.
>
>>I for one have actually learned much about the registry just by "looking"
>>at it. It does make sense, if you take time to see how it's laid out.
>
>I know. I discovered how the NT password hashes are stored
>in the SAM by "looking" at it. The problem with the registry
>is that default values usually aren't stored there at all.
>
>Thus not seeing a key or value doesn't tell you anything
>about if a particular key or value actually might exist,
>or what potential keys may be possible. That's what I meant
>by "undocumented".
>
>>I think you misunderstand my point. I'm referring to UNIX, in reference
>>to "why Apache has 50% of the market". Commercial Web servers for UNIX
>>cost quite a bit of money, last I checked.
>
>Well Apache has 50% of the *total* Web server
>market, not just UNIX only.
>
>>: The second edition covers NT 4.0. I ask again, what
>>: setting do *you* tune to make NT use more than 300MB for
>>: file cache memory.
>
HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\Session
Manager\LargeSystemCache:REG_DWORD=1
Enables caching to physical RAM -4MB.
>
>>NO! I'm sorry, but since you failed in your quest for the sacred
>>OracleOfTheNTRegistry(tm)(R)(C), you must give pennance by serving
>>20,000 users with a Timex Sinclair 1000. :-D
>
>If there was a document anywhere outside of Microsoft
>documenting it, then I'd like to know about it....
>
Check out "Windows NT Registry, A Settings Reference" by Sandra Osborne.
New Riders 1998
I've found it to be a useful reference on registry tweaking.
>
>>Seriously though, I personally don't know how to go about doing that. I
>>was just suggesting that there might be a way to do so under WindowsNT
>>v4.0, which perhaps you or I had not been made aware of. There
>>very well may not be a way... I honestly don't know.
>
>I'm coming to the conclusion that this *is* a hard
>limit in WinNT4.x. I'd be glad to be proven wrong
>though.
>
>Regards,
>
> Jeremy Allison.
Martin Ozolins
Jeremy Allison wrote in message ...
>
>>HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\Session
>>Manager\LargeSystemCache:REG_DWORD=1
>>Enables caching to physical RAM -4MB.
>
>as I'm back at work and have access to the machine
>and all my NT reference books again. It does sound
>familiar though.... hmmm. Are you sure this isn't
>the value that is set when you click the "tune NT
>as a file server" dialog ? I ask as it doesn't
>
Yes. it is, I don't see a method of controlling that specific parameter,
because NT dynamically assigns these values. Theoretically, you should be
able to run the FS Cache up to 508MB given 512MB RAM.
>
>contain a value in MB which I would have expected for a
>setting to tune the amount of cache used, rather
>than a boolean which is what this one seems to be.
>
>>Check out "Windows NT Registry, A Settings Reference" by Sandra Osborne.
>>New Riders 1998
>>I've found it to be a useful reference on registry tweaking.
>
>
>More when I've had time to run some more tests....
>
>Regards,
>
> Jeremy Allison.
Jeremy Allison
>HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\Session
>Manager\LargeSystemCache:REG_DWORD=1
>Enables caching to physical RAM -4MB.
Thanks a *lot* for that info. I'll check this out as soon
as I'm back at work and have access to the machine
and all my NT reference books again. It does sound
familiar though.... hmmm. Are you sure this isn't
the value that is set when you click the "tune NT
as a file server" dialog ? I ask as it doesn't
Jeremy Allison
>Jeremy Allison wrote in message ...
>>"Martin Ozolins" <Martin....@cwusa.com> writes:
>>
>>>HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\Session
>>>Manager\LargeSystemCache:REG_DWORD=1
>>>Enables caching to physical RAM -4MB.
>>
>>Thanks a *lot* for that info. I'll check this out as soon
>>as I'm back at work and have access to the machine
>>and all my NT reference books again. It does sound
>>familiar though.... hmmm. Are you sure this isn't
>>the value that is set when you click the "tune NT
>>as a file server" dialog ? I ask as it doesn't
>>
>Yes. it is, I don't see a method of controlling that specific parameter,
>because NT dynamically assigns these values. Theoretically, you should be
>able to run the FS Cache up to 508MB given 512MB RAM.
Ah. Therein lies the problem. I have *2GB* of RAM
and I'd rather like NT to use more of it than 512Mb
for file caching.
On Linux, I was able to set the server to use 1.8GB
for file system cache and I was hoping to be able
to do the same on NT. It doesn't seem like NT is able
to make use of all that memory when all you want
is file serving (which is why it ended up hitting
the disk much sooner than the Linux box and hence
going much slower).
Regards,
Jeremy Allison.
Timothy Kelley
wrote:
>I don't understand what you are saying here. No, you _can_ have
>NT fixed. Microsoft issues Service Packs to fix bugs found in
>NT's source code. If the problem is with the PC's hardware,
>you could send it back to the manufacturer.
Um, no, not by a long shot. With the service packs you get new
features, hence more bugs. Why can't I get the bug fixes and not the
features? One word. Money.
>In my experience with NT and Linux, you can expect to have
>far more problems with Linux than NT. Linux is like one of
>those run-down houses for sale called a "Handyman's Dream"
>by the real estate company. You can make it into something
>decent, but you may have to do the work yourself(by using the
>Linux source code).
Huh? What are you talking about? I've done all sorts of useful
things in linux without even looking at the source code. Linux is
user driven, NT is money driven. I have no problems whatsoever using
linux, every day, and on my desktop. What are you using, linux .03 or
something?
Mike Cantrell
> If this is so, then how is it that WindowsNT OEMs are able to provide
> technical support for their products? Do you think for one minute that
> COMPAQ, DELL, or HP do not have a plethora of information about WindowsNT
> the case, then OEMs wouldn't exist in the first place.
Trust me, they don't. As far as DELL goes, I spent half my time with
them explaining to *them* how the OS is supposed to work. Based on what
what my clients have received from HP and Compaq, I don't have much
confidence in them either.
The only vendor that I know that's made code-level changes to NT is
Citrix, and their tech-support, which costs a couple of grand to get
access to at all, is manned largely by fucking idiots who don't know
what products their company produces or what those products are supposed
to do.
>
> An OEM version of Microsoft software is that which is completely
> custom-tailored by an OEM, to suit their make and model of machines. How
> do you think OEMs can exist without knowing how to go about customizing
> the OS for themselves?
I can't speak for the others, but DELL's mods to the OS do not qualify
as customizations. Yeah, they call a little program that installs a
couple of apps after the fact, and might stick an icon in here and
there, but there is no code-level difference between NT that I buy from
a distributor, off the shelf, or from DELL. It's cosmetic changes,
that's all. I know, I've done it. That NT CD that DELL ships contains
none of their mods, it's *identical* to a CD that comes from MS except
for that DELL logo silk-screened on to the top. All of DELL's add-ins
are stored in a partition on the drive and loaded after the OS is
installed.
--Mike
Mike Cantrell
Stephen Edwards wrote:
> You don't have to be a genius to understand WindowsNT either. You just
> have to be willing to spend a little more time learning
> TheMicrosoftWay(tm). They both make perfect sense to me, and if __I__ can
> understand both WindowsNT and UNIX, then __ANYONE__ can.
>
> In short, WindowsNT is a different "way of thinking" than UNIX is.
>
I don't see how anyone who claims to understand UNIX can possibly assert
that the MS Way makes perfect sense. Don't get me wrong, MS provides me
with a *very* conmfortable living. But so many "design decisions" that
MS makes just strike me as being absolutely brain dead. I've used UNIX,
I play with Linux, and I make my living administering NT. And there are
days the I would happily kill someone to get some of the functionality
of UNIX in NT.
--Mike
troth
work equally well with NT.
Mark Hamstra
> Nothing about linux makes Beowolf more scaleable. The concepts of Beowolf
> work equally well with NT.
No, they don't. The pile of PCs approach of Beowulf demands a highly
reliable OS, else you'll feel like one of the vacuum tube replacing
drones of the ENIAC era, except now you'll be rebooting Windows boxes
instead of replacing failed tubes.
Additionally, Beowulf requires high-performance network interfaces,
both in terms of latency and throughput. NT can match neither
specification of the Linux TCP/IP stack and the best Linux interface
drivers, many of which have been written and/or fine-tuned by Beowulf
developers --something that can't be done with closed NT source code.
Beyond that, there are a number of very annoying "features" of NT
that make it of limited appeal in a Beowulf-style environment. Not
the least of which is that MPI, PVM, and related tools are primarily
Unix-oriented, meaning that an NT Beowulf user would first need to
spend a great deal of time porting the necessary toolsets. The need
for multiple display heads and keyboards (or switch boxes) and the
lack of remote administration tools on NT also detract from the
appeal of an NT Beowulf.
None of these are really fundamental kernel technologies that are
unique to Linux, but they do add up to real positives for Linux in
Beowulf clusters and real negatives for any similar use of NT --i.e.,
much more hassle for a good deal less performance when using an NT
Beowulf vs. Linux Beowulf.
--
Mark Hamstra
Bentley Systems, Inc.
Martin Ozolins
Jeremy Allison wrote in message ...
>"Martin Ozolins" <Martin....@cwusa.com> writes:
>
>
>>Jeremy Allison wrote in message ...
>>>"Martin Ozolins" <Martin....@cwusa.com> writes:
>>>
>>>>HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\Session
>>>>Manager\LargeSystemCache:REG_DWORD=1
>>>>Enables caching to physical RAM -4MB.
>>>
>>>Thanks a *lot* for that info. I'll check this out as soon
>>>as I'm back at work and have access to the machine
>>>and all my NT reference books again. It does sound
>>>familiar though.... hmmm. Are you sure this isn't
>>>the value that is set when you click the "tune NT
>>>as a file server" dialog ? I ask as it doesn't
>>>
>>Yes. it is, I don't see a method of controlling that specific parameter,
>>because NT dynamically assigns these values. Theoretically, you should be
>>able to run the FS Cache up to 508MB given 512MB RAM.
>
RAM-4MB. You do the math.
Sean Burke
>
> "Boris" <boris...@pleasemovil.com> writes:
>
> > LINUX crowd is full of crap-producing, self-respected "gurus" (LINUX is just one crap they
> > produce, SAMBA is another).
>
> It certainly is odd then how that "crap" Samba running on
> that "crap" OS, Linux, soundly outperforms NT's SMB filesharing.
>
> It is olso very odd that the "self-respected" Jeremy Allison
> took no public affront to your unwarranted indictment of his
> NT knowledge, preferring instead a very low-key approach that
> lets the performance numbers speak for themselves.
Perhaps Jeremy knows that Boris is one of the oldest
trolls in comp.os.*.advocacy. He seems to have toned
down the crude language (believe it or not! :^)
-SEan
Jeremy Allison
>>>Yes. it is, I don't see a method of controlling that specific parameter,
>>>because NT dynamically assigns these values. Theoretically, you should be
>>>able to run the FS Cache up to 508MB given 512MB RAM.
>>
>The example given is just that an example. The actual assignment is
>RAM-4MB. You do the math.
Oh, I see. Sorry for being dumb :-(. The reason I
misunderstood is that 512MB is very close to the
value that the Windows NT Internals book claims
is the file system cache limit for NT.
I'll ensure this registry value is set and bang on NT some
more. I'll let you know if I see it using more than 300Mb
for file cache.
Cheers,
Jeremy Allison.
troth
Everybody states that you need to reboot your servers constantly or you get
bsod's or other bs. I've never had to reboot an nt server (I have 10) in
over two years at this point and none of my server are simple file and print
servers!
The company that needs to be concerned with Linux is not MS but more likely
Novell.
Both MPI and PVM are available for NT.
Jerry
HEARD OF that are anything like that reliable. I would have thought that
Microsoft would be interested in using your axample for publicity
purposes........unless you are being economical with the truth that
is....
Did'nt reboot in 2 years - that presumably means no re-configuration
then ? Can't be under much load either - perhaps you are not such a good
example....
Mark Hamstra
> I'd really like to verify some of the comments made about NT reliability.
> Everybody states that you need to reboot your servers constantly or you get
> bsod's or other bs. I've never had to reboot an nt server (I have 10) in
> over two years at this point and none of my server are simple file and print
> servers!
>
> The company that needs to be concerned with Linux is not MS but more likely
> Novell.
It's not an either/or: any company that is relying upon
sales of a workgroup class server OS that is doing
primarily file and print sharing needs to worry. Linux
can do a great deal more than just file and print sharing,
but it does those basic tasks better than either NT or
Netware, so both Novell and Microsoft should be worried.
> Both MPI and PVM are available for NT.
Yes, they are; but the vast majority of developers using
them and developing additional tools for them are doing
so in a Unix environment. It is not impossible to use
NT in a Beowulf-like configuration, it is just a lot less
convenient to do so than to use Linux, and the performance
wouldn't be as good.
> the least of which is that MPI, PVM, and related tools are primarily
> > Unix-oriented, meaning that an NT Beowulf user would first need to
> > spend a great deal of time porting the necessary toolsets.
--
jesus
> Everybody states that you need to reboot your servers constantly or you get
> bsod's or other bs. I've never had to reboot an nt server (I have 10) in
> over two years at this point and none of my server are simple file and print
> servers!
you must be the luckiest sob in the galaxy
why not use that luck and buy a lottery ticket.
at least you can make money from it. whereas,
only ms really makes money off of nt.
Leslie Mikesell
troth <tr...@or.blm.gov> wrote:
>I'd really like to verify some of the comments made about NT reliability.
>Everybody states that you need to reboot your servers constantly or you get
>bsod's or other bs. I've never had to reboot an nt server (I have 10) in
>over two years at this point and none of my server are simple file and print
>servers!
How do you do something as simple as adding an external modem without
rebooting?
Les Mikesell
l...@mcs.com
Martin Ozolins
rebooted either, except for scheduled maint. which occurs on my UNIX boxes
as well. Might be that cause some can't figure it out it must be at fault.
troth wrote in message ...
>I'd really like to verify some of the comments made about NT reliability.
>Everybody states that you need to reboot your servers constantly or you get
>bsod's or other bs. I've never had to reboot an nt server (I have 10) in
>over two years at this point and none of my server are simple file and
print
>servers!
>
>The company that needs to be concerned with Linux is not MS but more likely
>Novell.
>
>Both MPI and PVM are available for NT.
Martin Ozolins
:-)
Bloody Viking wrote in message <7f11rm$a1r$4...@hirame.wwa.com>...
>In comp.os.linux.advocacy Stephen Edwards <ja20...@primenet.com> wrote:
>
>: But you're right, there are many a UNIX user who would rather just
abandon
>: WindowsNT, and conclude that "it's crap", rather than understand what it
>: is exactly that they are doing wrong. I've made a few BSD users eat
their
>: words in my time. Not many, but a few. :)
>
>You really need to see the almost religious fervour with just BOOTING
>Linux. It seems there are several ways to boot it. One way is you can boot
>it from DOS, much like Novell Netware, and the other method is this boot
>loader that except for select BIOSes doesn't work and sometimes trashes
>hard drives.
>
>I use the Loadlin (boot from a DOS) method, but the zealots insist that
>I'm stupid becuse I can't get the One True Way (LILO) to work due to flaky
>BIOSes. I did find out the Big Secret. You only buy motherboards from
>exactly one manufacturer. (American Megatrends) However, I found out only
>after someone posted a link and I found out by using it.
>
>I'm a Linux fan myself, but when I see a problem with it, I point it out.
>But the zealots don't want to hear there is any problem at all. In short,
>these LILO fanatics are like religious fundimentalists. The use of any
>other boot method constitutes Heresy if not Blasphemy to the Linux
>fundimentalists.
>
>Maybe someone needs to newgroup alt.religion.linux or
>alt.religion.operating-system.* for the OS Warz and the occasional boot
>loader wars.
>
>--
>CAUTION: Email Spam Killer in use. Leave this line in your reply! 152680
> First Law of Economics: You can't sell product to people without money.
>
>3869387 bytes of spam mail deleted. http://www.wwa.com/~nospam/
matt
>
> matt wrote:
>
> > You are right Linux users have not mastered NT neither has even the best NT
> > admin. We can't. NT is closed source. We can not go in and look at the NT
> > source code to figure out EXACTLY what happens when you issue a command. It
> > is not that we are so stupid that we can not understand NT, It is that we
> > refuse to be stay at the level of stupidity required to be happy with NT.
>
> If you must look at the source code to use any operating system or
> application software, there is something seriously wrong with the
> software's interface or the interface's documentation. You do not
> need to see the schematics for your television to change the channel.
> You do not have to see the design of the engine to drive your car.
> If you are not a developer of the operating system, you should never
> be required to see the source code. Period.
Ummm, the post I responded to was saying that Linux users did not
understand NT, After someone posted something that Linux users were so
stupid they did not understand NT... Since this is about UNDERSTANDING
HOW SOMETHING WORKS... a person who can read and understand electronics
UNDERSTANDS how a TV works better than the person who does not. The car
is the best analogy of all... Sure you can drive a car without
understanding how the engine works. However, if you understand how the
engine works you'll understand what it takes to keep it running at Peak
Performance (TM). If you don't know how a car works, how do you know if
you're getting ripped off by the mechanic?? The more you understand how
a car works, the more you can get out of the car (better performance,
longer live, etc). The more you undersand how a OS works (including how
the source code is writen), The more you can get out of the OS
Now, if you're only smart enough to change a channel on a TV or drive a
car without understanding how an internal combustion engine works, stay
with NT. I was written for you.
If you're smart enough and want to lean how you're computer works so you
can keep it running at Peak Performance (TM) use Linux. At least I know
enough to know when I am getting ripped off by paying for over priced
software and paying for useless customer support...
Bloody Viking
Bloody Viking
: Heretic!!!
: :-)
Seriously, you should come and see the hate posted by LILO fundies, and
see my retaliation with equal flamage. It's all becuse of a computer
meltdown with my attempt to use LILO and my getting pissed off about
trashing hard drives thanks to a bad BIOS. I didn't find out that part
only until the damage was done.
If I ever want to fuck up someone's computer, all I need is a Red Hat CD.
And I can fuck it up to a point he'll need a new hard drive. The ultimate
computer revenge. It's even nastier than a mere format. All you have to do
is just give them the album and tell them how great Linux is, and let LILO
do the deed. And by going to Cheapbytes, the dirty deed can be done dirt
cheap.
:)hehehe
Chris Costello
> I use the Loadlin (boot from a DOS) method, but the zealots insist that
> I'm stupid becuse I can't get the One True Way (LILO) to work due to flaky
> BIOSes. I did find out the Big Secret. You only buy motherboards from
> exactly one manufacturer. (American Megatrends) However, I found out only
> after someone posted a link and I found out by using it.
Use what works for *you* not comp.os.linux.advocacy.
Anthony Ord
<tr...@or.blm.gov> wrote:
>I'd really like to verify some of the comments made about NT reliability.
>Everybody states that you need to reboot your servers constantly or you get
>bsod's or other bs. I've never had to reboot an nt server (I have 10) in
>over two years at this point and none of my server are simple file and print
>servers!
Simplicity, or otherwise of the functions seems not to
matter with NT. It appears to be the load.
If you have a web server serving static content, then the
load should be quite low (dynamic is another kettle of
fish). File servers are usually more highly loaded, but that
depends on the CPU/disk IO mix.
>The company that needs to be concerned with Linux is not MS but more likely
>Novell.
Novell have released Unixware for Linux - available at
Caldera's site AFAIR.
>Both MPI and PVM are available for NT.
> the least of which is that MPI, PVM, and related tools are primarily
>> Unix-oriented, meaning that an NT Beowulf user would first need to
>> spend a great deal of time porting the necessary toolsets.
Regards
Anthony
--
-----------------------------------------
| And when our worlds |
| They fall apart |
| When the walls come tumbling in |
| Though we may deserve it |
| It will be worth it - Depeche Mode |
-----------------------------------------
Craig Kelley
> If I ever want to fuck up someone's computer, all I need is a Red Hat CD.
> And I can fuck it up to a point he'll need a new hard drive. The ultimate
> computer revenge. It's even nastier than a mere format. All you have to do
> is just give them the album and tell them how great Linux is, and let LILO
> do the deed. And by going to Cheapbytes, the dirty deed can be done dirt
> cheap.
The Disk Druid is an unforgiving mistress.
Of course, all *I* need to screw up a system is a large electro-magnet.
:> This HP Vectra is going to get it soon, if it doesn't start behaving!
--
The wheel is turning but the hamster is dead.
Craig Kelley -- kell...@isu.edu
http://www.isu.edu/~kellcrai finger i...@inconnu.isu.edu for PGP block
Timothy Kelley
"David J. Owens" wrote:
>
> matt wrote:
>
> > You are right Linux users have not mastered NT neither has even the best NT
> > admin. We can't. NT is closed source. We can not go in and look at the NT
> > source code to figure out EXACTLY what happens when you issue a command. It
> > is not that we are so stupid that we can not understand NT, It is that we
> > refuse to be stay at the level of stupidity required to be happy with NT.
> If you must look at the source code to use any operating system or
> application software, there is something seriously wrong with the
> software's interface or the interface's documentation. You do not
> need to see the schematics for your television to change the channel.
> You do not have to see the design of the engine to drive your car.
> If you are not a developer of the operating system, you should never
> be required to see the source code. Period.
In this case it isn't the source code. While the open source code
does imapact all users in a very strong way, it does so indirectly.
In the case of a non coder (eg. a sysadmin or network engineer), Linux
is still far easier to work with than NT. Because it is open source
GPL, it is end user driven. Because is end user driven, it has things
end users want (well the end users of linux are pretty geeky). As a
result of this, linux is generally very easy to work with.
A good example is just how verbose it is when you're trouble shooting
problems in linux. NT is basically silent in comaprison, forcing one
into neanderthal-like trial and error procedures most of the time.
Another example is the registry vs. text configuration files. No end
user would willingly accept something like the windows registry. It's
an abomination.
Yet another is the concept of following standards, so everyone can
communicate. Micros~1 still doesn't like this idea.
Steve Mading
>I use the Loadlin (boot from a DOS) method, but the zealots insist that
>I'm stupid becuse I can't get the One True Way (LILO) to work due to flaky
>BIOSes. I did find out the Big Secret. You only buy motherboards from
>exactly one manufacturer. (American Megatrends) However, I found out only
>after someone posted a link and I found out by using it.
Liar. Others have already posted that their LILO works fine under other
BIOSes other than AMI. I am using it with Pheonix, for example. (Please note,
I am not saying that you are lying about your bad BIOS. I believe you there. I
am saying that you are lying when you state that only AMI works.
(Terminology note, the BIOS is not the motherboard. Be careful how ou phrase
things.). )
Ciaran Dunn
Boris wrote in message <371097f6$0$2...@nntp1.ba.best.com>...
>LINUX crowd is full of crap-producing, self-respected "gurus" (LINUX is
just one crap they
>produce, SAMBA is another).
>
It is for such well formed and thought out responses from people like Boris
that I read this newsgroup :)
I was a Linux user but now, since Boris's well thought out, lucent and
convincing argument, I have come to realise that it is "crap". Thanks
Boris... youve saved me valuable time... I now realise that NT is the one
true OS for all applications... Im off to format my hard disks and buy a
copy of NT4.
Cheers,
Ciaran
Volker Dittmar
>
> I'd really like to verify some of the comments made about NT reliability.
> Everybody states that you need to reboot your servers constantly or you get
> bsod's or other bs. I've never had to reboot an nt server (I have 10) in
> over two years at this point and none of my server are simple file and print
> servers!
Take a look at SAP/R3. Load the transactions of a few accounts with more than
100,000 transactions. If you have 64MB, NT will crash after 10 minutes. If you
have 128MB, NT will crash after 20 minutes (BSOD). It works fine with other OS
(and, BTW, it doesn't matter on which OS the database runs - it is the client
which crashes).
Yes, its THAT easy to crash NT.
As a server: if you run M$ SQL Server, NT will crash if you push it hard
enough (one of the reasons why you won't find TPC-D benchmark results for
NT).
As I said before: if your NT didn't crash recently, you haven't pushed it
hard enough.
Ciao,
Volker
Revenant
>I know a lot of end users that prefer the command line to the gui.
>Which is easier to say on a phone call, or explain:
>
>a. "ok, now I want you to type ls, and press the enter key."
>or
>b. "now, I want you to click on the task bar. that's the little
bar
>on the bottom of the screen. oh, you don't have one on the
>bottom of the screen? Well, check to see if it is on the top or
>on the sides of the screen. It has a little clock on it, unless you
>checked it off. No mam, I don't know what color it is."
>"Ok, you found the taskbar. great, now move the mouse over
> it, and click on it with the right mouse button. You should get
>a little menu that. Oh a menu, that's the list of stuff. One of
>those things will say "minimize all windows". Click on that.
You have a good point re: the simplicity (actually, directness) of
using a command line rather than a GUI. But you shot yourself
in the foot with the example you chose. Try:
"ok, now I want you to hold down the key with the Windows Logo and
press 'M'".
Conversely if you were after a specific directory, try:
"ok, now I want you to hold down the key with the Windows Logo and
press 'R'. Now, in the box that appears type <directory name>"...
---------------------------------------------------
This message contains personal opinion and
probably does not represent the opinion of my
employer.
Sang K. Choe
wrote:
You may also want to test it on a SCSI based system (I recall reading
the spec and you were using a IDE based system no?). I know what NT
really doesn't care for IDE (even the new fangled UltraDMA/66) based
disks.
There's a benchmark run by folk over at Mindcraft which compared
RedHat 5.2 with a 2.2.2 kernel upgrade running Samba 2.0.3 against NT
4.0. It's quite interesting...(read: Linux + Samba didn't look too
good...)
http://www.mindcraft.com/whitepapers/nts4rhlinux.html
(This ought throw a bit more wood on the fire... 8*)
-- Sang.
------------------------------------------------------------
My opinions are my own...
matt
> "ok, now I want you to hold down the key with the Windows Logo and
> press 'M'".
>
> Conversely if you were after a specific directory, try:
>
> "ok, now I want you to hold down the key with the Windows Logo and
> press 'R'. Now, in the box that appears type <directory name>"...
This, of course assumes that the user has Paid Even More Money (TM)
upgrading his/her system by buing a new keybord... If they have an old
key bord, Oh well. .
Christopher Smith
matt <cyb...@pacbell.net> wrote in message
news:37157DCA...@pacbell.net...
>
> > "ok, now I want you to hold down the key with the Windows Logo and
> > press 'M'".
> >
> > Conversely if you were after a specific directory, try:
> >
> > "ok, now I want you to hold down the key with the Windows Logo and
> > press 'R'. Now, in the box that appears type <directory name>"...
>
> This, of course assumes that the user has Paid Even More Money (TM)
> upgrading his/her system by buing a new keybord... If they have an old
> key bord, Oh well. .
Umm, if anything they would have had Paid Even Less Money (TM) for a
keyboard with a windows key.
Des Herriott
> LINUX crowd is full of crap-producing, self-respected "gurus" (LINUX is just one crap they
> produce, SAMBA is another).
1) Boris accuses Jeremy Allison of being ignorant about NT.
2) Mark points out to Boris that Jeremy is actually one of the Samba authors.
3) Boris responds with his usual toilet-mouthed comments, presumably because
he doesn't have a good answer.
Boris, you truly are your own worst enemy. I'd give up advocating NT right
now if I were you, because you really don't know how to do it.
--
Des Herriott, Oracle Corporation UK Ltd.
dher...@uk.oracle.com
- speaking for myself, not my employer.
Scott Campbell
for those of us who do not (My Thinkpad 380XD does not have a windows
key).
scott
Revenant wrote:
>
> Todd Bandrowsky wrote in message <370f...@news.deniz.com>...
> >I know a lot of end users that prefer the command line to the gui.
> >Which is easier to say on a phone call, or explain:
>
> You have a good point re: the simplicity (actually, directness) of
> using a command line rather than a GUI. But you shot yourself
> in the foot with the example you chose. Try:
>
Craig Kelley
> There's a benchmark run by folk over at Mindcraft which compared
> RedHat 5.2 with a 2.2.2 kernel upgrade running Samba 2.0.3 against NT
> 4.0. It's quite interesting...(read: Linux + Samba didn't look too
> good...)
>
> http://www.mindcraft.com/whitepapers/nts4rhlinux.html
http://www.lwn.net/1999/features/MindCraft1.0.phtml