Linux hype ....
Unknown
shit.....
Im running Fedora core 3/4 and Debian across 8 machines and its all
starting to piss me off to the point im considering moving back to
windows.
Some examples:
1) Often when I yum update something breaks - often something difficult to
fix
2) Updates causes firefox to lose its java plugin
3) Newer java runtime causes firefox not start, start from a console for
NO debug output whatsoever - now thats well designed for "cutting edge
browser"
4) Pan misses and duplicates characters as I type this.
5) Open-office crashes and is not backward compatible with a lot of xls
files.
6) NFS and evolution interact - the "stateless server" NFS server causes
fucking unhelpful error messages about lock ioctl or some such in
evolution. Restarting the client doesnt fix it - you have to restart the
server (or at least NFS service).
7) Gnome now suffers window lockage just like windows. Open a dialog in
one application blocks the opening of dialog in another.
8) SSH setup isn't consistent across linux distros - some do X11
forwarding others dont - some set the DISPLAY variable others dont ...
very fucking annoying if you have lots of machines to admin.
9) The only thing constant about administering apache across distros is
the name - every other fucking thing has a different path/filename...
10) Evolution is the best visual client - but you still cant drag and drop
attachments from within KDE .... desktop integration pha !
Im I alone in having these day to day problems - im getting the impression
that things are getting worse not better as the months tick by.
Jason Clifford
> Im running Fedora core 3/4 and Debian across 8 machines and its all
> starting to piss me off to the point im considering moving back to
> windows.
>
> 8) SSH setup isn't consistent across linux distros - some do X11
> forwarding others dont - some set the DISPLAY variable others dont ...
> very fucking annoying if you have lots of machines to admin.
So standardise upon a single distribution or understand that software like
OpenSSH isn't tied to a distribtion and you can configure it as you like.
Look in /etc/ssh/
> 9) The only thing constant about administering apache across distros is
> the name - every other fucking thing has a different path/filename...
Yes because it's basically assumed that if you select a distribution you
intend to use that distributions way of doing things. If you want a simple
environment then keep it simple and standardise on one distribution.
If you want one that is really good for desktop try Ubuntu. For servers
Debian is the way to go.
> 10) Evolution is the best visual client - but you still cant drag and drop
> attachments from within KDE .... desktop integration pha !
If you want desktop integration use a KDE application such as KMail. I
hope you are not one of those who tried to send documents "within" an
email rather than as proper attachments.
Jason
--
UKFSN.ORG Finance Free Software while you surf the 'net
http://www.ukfsn.org/ 2Mb ADSL Broadband from just £14.98 / month
Tony van der Hoff
<pan.2005.09.26....@way.cc>
> Im coming to the conclusion that the Linux desktop hype is a sack of
> shit.....
>
> Im running Fedora core 3/4 and Debian across 8 machines and its all
> starting to piss me off to the point im considering moving back to
> windows.
>
> Some examples:
>
[snip]
>
ISTM when you run into problems like this, there are a number of strategies
to make you feel better:
1. Rant. Whilst possibly good for the soul, it doesn't do much towards
fixing the problem.
2. Try to analyse and understand the problem, possibly with the help of a
newsgroup such as this. Can be highly frustrating in the short term, but
ultimately very rewarding when you achieve a stable system which does what
you require.
3. Move to Windows, or, for that matter any other OS, and experience a new
set of problems to solve. Rewarding in the short term, but just as
frustrating in the longer term. Move again.
4. Abandon computers, and take up crochet. Peaceful existence.
Note that these options are mutually exclusive, and there's no magic
available to help you out.
Now personally, I would opt for (2), but I guess that's just my
investigative mind at work. Your choice of (1) is probably not a good start
in getting people here to help you out, but that's maybe not what you're
looking for. Options (3) or (4) are possibly your best course of action,
given your current mindset.
> Im I alone in having these day to day problems - im getting the impression
> that things are getting worse not better as the months tick by.
>
You're probably not alone; I guess there are many people out there who are
simply incapable of understanding the actions necessary to get a system
working properly. Nonetheless, there are *millions* of highly-satisfied
gnu-linux users, using whatever distro, which they have honed and fine-tuned
to their satisfaction.
This may be what you're looking for, and there are plenty of people here who
will respond positively to a genuine cry for help, once you've tried to help
yourself. All you have to do is start with the right state of mind, and
nobody can or will help you with that.
--
Tony van der Hoff | mailto:to...@vanderhoff.org
Buckinghamshire, England
Roger Leigh
Hash: SHA1
"Unknown" <no...@way.cc> writes:
[There's no penalty for using your real name, you know...]
> 2) Updates causes firefox to lose its java plugin
>
> 3) Newer java runtime causes firefox not start, start from a console for
> NO debug output whatsoever - now thats well designed for "cutting edge
> browser"
> 5) Open-office crashes and is not backward compatible with a lot of xls
> files.
These applications all fall into the "big bloated monster" category.
None of them do a particularly great job of integrating with the host
desktop environment, and try to do everything rather than delegate
anything to other libraries or programs, or in the case of Java, not
making any effort to fit nicely with the host operating system.
- --
Roger Leigh
Printing on GNU/Linux? http://gimp-print.sourceforge.net/
Debian GNU/Linux http://www.debian.org/
GPG Public Key: 0x25BFB848. Please sign and encrypt your mail.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.2 (GNU/Linux)
Comment: Processed by Mailcrypt 3.5.8+ <http://mailcrypt.sourceforge.net/>
iD8DBQFDODo+VcFcaSW/uEgRApM2AJ9lvx9vonawE2cbDTdavgJzyFMznACfc3V6
E7fcNYOtZ/SAjWo1tYYhJwA=
=Rmks
-----END PGP SIGNATURE-----
Unruh
>Im coming to the conclusion that the Linux desktop hype is a sack of
>shit.....
Sounds to me like you have problems with your hardware.
Also why in the world would you mix Debian and FC? YOu are trying hard to
give yourself problems.
>Im running Fedora core 3/4 and Debian across 8 machines and its all
>starting to piss me off to the point im considering moving back to
>windows.
>Some examples:
>1) Often when I yum update something breaks - often something difficult to
>fix
I use Mandrake and urpmi. And updates are trivially easy (across 10
machines) and have not broken anything. But I cannot say about yum. It may
be a problem with yum or it may be a problem with the user of yum.
>2) Updates causes firefox to lose its java plugin
On which system?
>3) Newer java runtime causes firefox not start, start from a console for
>NO debug output whatsoever - now thats well designed for "cutting edge
>browser"
?? No idea what you are saying here.
>4) Pan misses and duplicates characters as I type this.
Again, noone else has this problem. On which system do you have this
problem?
>5) Open-office crashes and is not backward compatible with a lot of xls
>files.
Again, on which system? And where did you create those xls files?
>6) NFS and evolution interact - the "stateless server" NFS server causes
>fucking unhelpful error messages about lock ioctl or some such in
>evolution. Restarting the client doesnt fix it - you have to restart the
>server (or at least NFS service).
AGain.
>7) Gnome now suffers window lockage just like windows. Open a dialog in
>one application blocks the opening of dialog in another.
???
>
>8) SSH setup isn't consistent across linux distros - some do X11
>forwarding others dont - some set the DISPLAY variable others dont ...
>very fucking annoying if you have lots of machines to admin.
Uh, yes, different distros may well have different default sshd.config
files. Make them the same.
>9) The only thing constant about administering apache across distros is
>the name - every other fucking thing has a different path/filename...
So why in the world are you running different distros? That is a very very
silly thing to do.
>10) Evolution is the best visual client - but you still cant drag and drop
>attachments from within KDE .... desktop integration pha !
>Im I alone in having these day to day problems - im getting the impression
>that things are getting worse not better as the months tick by.
You seem to be relatively alone.
A) Run a common distribution. That will solve a bunch of your problems.
B) Check out your hardware.
Ken Parkes
> 1) Often when I yum update something breaks .....
> 4) Pan misses and duplicates characters as I type this.
> 5) Open-office crashes ......
>
> 7) Gnome now suffers window lockage just like windows.
...and, for me, distro installs hung.
Solution was money; I had a RAM problem.
Ken.
Paul F. Johnson
On Mon, 2005-09-26 at 17:30 +0000, Unknown wrote:
> Im coming to the conclusion that the Linux desktop hype is a sack of
> shit.....
Nope, can't say I'd agree...
> Im running Fedora core 3/4 and Debian across 8 machines and its all
> starting to piss me off to the point im considering moving back to
> windows.
You're running (effectively) 3 distros there and are expecting
everything to run the same? While FC3 and 4 should be roughly the same,
Debian probably won't be.
> Some examples:
>
> 1) Often when I yum update something breaks - often something difficult to
> fix
Such as? Have you reported it back to whichever distro it came from? I
can't say that I've noticed anything much break with either FC3 or 4.
> 2) Updates causes firefox to lose its java plugin
That is annoying. I'll give you that.
> 3) Newer java runtime causes firefox not start, start from a console for
> NO debug output whatsoever - now thats well designed for "cutting edge
> browser"
Nope. Not seen that one. Which version of FF and which JVM are you
using? More over, is it happening with all three distros?
> 4) Pan misses and duplicates characters as I type this.
Not used Pan in ages. Thunderbird is a far better usenet agent. That
said, I just use Evolution and be done with it.
> 5) Open-office crashes and is not backward compatible with a lot of xls
> files.
Version? Distro? You'll find that Excel doesn't really understand being
backward compatible. Infact, half the time it'll reject files *it*
created!
> 7) Gnome now suffers window lockage just like windows. Open a dialog in
> one application blocks the opening of dialog in another.
Nope. Never seen that. I think you're just having bad luck.
> 8) SSH setup isn't consistent across linux distros - some do X11
> forwarding others dont - some set the DISPLAY variable others dont ...
> very fucking annoying if you have lots of machines to admin.
OpenSSH is not distro specific - you need to understand that you may
need to alter the config file to enable forwarding.
> 9) The only thing constant about administering apache across distros is
> the name - every other fucking thing has a different path/filename...
No, they should have the same filename. Where that file is stored though
is distro specific.
> 10) Evolution is the best visual client - but you still cant drag and drop
> attachments from within KDE .... desktop integration pha !
Um, not with you there. Are you trying to write an email and embed a
jpeg in or a word processed file?
> Im I alone in having these day to day problems - im getting the impression
> that things are getting worse not better as the months tick by.
Right, you need to do the following
1. Take a deep breath
2. Standardise on a *single* distro [1]
3. Learn how to *use* the distro
4. Sit back and enjoy countless hours of trouble free computing.
[1] I've been on the RedHat/Fedora treadmill for goodness knows how long
now and can't be arsed with using the stable versions so have 5 machines
all on rawhide. I expect things to break, but they don't. I think the
last time I used the "stable" version was RH 7.1. Now, think about it. I
am still using a single distro and know where things are put by the
distro. If I move to a Slackware box, it takes a while to re-aclimitise,
but it's not hard. My nightmare is being stuck on a Debian box (but
then, I never have liked Debian!).
Most of your problems seem to be down to (what looks to be) a lack of
patience and the idea that ever Linux distro has everything in the same
place.
TTFN
Paul
--
"Duirt me leat go raibh me breoite." - T.M.
Neil Ellwood
> Im coming to the conclusion that the Linux desktop hype is a sack of
> shit.....
>
> Im running Fedora core 3/4 and Debian across 8 machines and its all
> starting to piss me off to the point im considering moving back to
> windows.
You should have got one m/c running reliably first.
>
> Some examples:
>
> 1) Often when I yum update something breaks - often something difficult to
> fix
Use google to find out what you are doing wrong.
>
> 2) Updates causes firefox to lose its java plugin
Has never happened to me (I have gone from man 7.2 right through to
Mandriva 2006 beta 2)
>
> 3) Newer java runtime causes firefox not start, start from a console for
> NO debug output whatsoever - now thats well designed for "cutting edge
> browser"
If your not happy with that got to Epiphany.
>
> 4) Pan misses and duplicates characters as I type this.
What have You done to cause this - it has never happened to me.
>
> 5) Open-office crashes and is not backward compatible with a lot of xls
> files.
Install it properly.
>
> 6) NFS and evolution interact - the "stateless server" NFS server causes
> fucking unhelpful error messages about lock ioctl or some such in
> evolution. Restarting the client doesnt fix it - you have to restart the
> server (or at least NFS service).
As with the other items above you haven't supplied ANY info.
>
> 7) Gnome now suffers window lockage just like windows. Open a dialog in
> one application blocks the opening of dialog in another.
The use KDE or Evolution or...........
>
> 8) SSH setup isn't consistent across linux distros - some do X11
> forwarding others dont - some set the DISPLAY variable others dont ...
> very fucking annoying if you have lots of machines to admin.
Don't get annoyed try Google to help yourself.
>
> 9) The only thing constant about administering apache across distros is
> the name - every other fucking thing has a different path/filename...
You haven't got a problem then.
>
> 10) Evolution is the best visual client - but you still cant drag and
> drop attachments from within KDE .... desktop integration pha !
Get programming skills going and do it.
>
> Im I alone in having these day to day problems - im getting the
> impression that things are getting worse not better as the months tick
> by.
YES
--
Neil
Delete delete to reply by email
Unknown
> On Mon, 26 Sep 2005 17:30:44 +0000, nope wrote:
>
>> Im coming to the conclusion that the Linux desktop hype is a sack of
>> shit.....
>>
>> Im running Fedora core 3/4 and Debian across 8 machines and its all
>> starting to piss me off to the point im considering moving back to
>> windows.
> You should have got one m/c running reliably first.
>>
>> Some examples:
>>
>> 1) Often when I yum update something breaks - often something difficult to
>> fix
> Use google to find out what you are doing wrong.
[[ Ahh I see ... running yum update makes ME wrong - not the fucking
packagers - or linux itself - or the new GCC, nahhh ... its me ...
google to fix it, wow thats an idea i've never had - thank you for your
words of wisdom.
>>
>> 2) Updates causes firefox to lose its java plugin
> Has never happened to me (I have gone from man 7.2 right through to
> Mandriva 2006 beta 2)
Your not running either of two distros i've mentioned, this makes yours
the comments worthless.
>>
>> 3) Newer java runtime causes firefox not start, start from a console for
>> NO debug output whatsoever - now thats well designed for "cutting edge
>> browser"
> If your not happy with that got to Epiphany.
>>
>> 4) Pan misses and duplicates characters as I type this.
> What have You done to cause this - it has never happened to me.
Who ho ...
>>
>> 5) Open-office crashes and is not backward compatible with a lot of xls
>> files.
> Install it properly.
>> 6) NFS and evolution interact - the "stateless server" NFS server causes
>> fucking unhelpful error messages about lock ioctl or some such in
>> evolution. Restarting the client doesnt fix it - you have to restart the
>> server (or at least NFS service).
> As with the other items above you haven't supplied ANY info.
[[I don't get ANY info - thats it ... a stupid error dialog picking up
email, something like "unable to ioctl lock" or similar - NOTHING more, no
filenames, no paths, no context, no error logs... Reboot the client and
its in the same state, restart NFS on the server it works. So NFS is
stateless, except for the locking thats stateful and buggy... wow .. much
betters than SMB...
>> 7) Gnome now suffers window lockage just like windows. Open a dialog in
>> one application blocks the opening of dialog in another.
> The use KDE or Evolution or...........
>>
>> 8) SSH setup isn't consistent across linux distros - some do X11
>> forwarding others dont - some set the DISPLAY variable others dont ...
>> very fucking annoying if you have lots of machines to admin.
> Don't get annoyed try Google to help yourself.
[[I can help myself - im not complaining I don't know how, im complaining
that stupid non-portable configurations make it time consuming and
tedious.
>> 9) The only thing constant about administering apache across distros is
>> the name - every other fucking thing has a different path/filename...
> You haven't got a problem then.
>>
>> 10) Evolution is the best visual client - but you still cant drag and
>> drop attachments from within KDE .... desktop integration pha !
> Get programming skills going and do it.
That has to be single most stupid thing any fuckwit has said so far. My
programming skills will enable me to steer the compatibility of a KDE
desktop and a gnome application will it? I steer the two projects and draw
up the standards do I ? Im sorry last time I looked I wasn't part of the
open desktop joke - you know the one that allows 3 clipboards that 1/2
work and drag and drop that doesn't work.....
>>
>> Im I alone in having these day to day problems - im getting the
>> impression that things are getting worse not better as the months tick
>> by.
> YES
Ahh silly me ... as KDE slides into bloat and GNOME tries to reinvent a
poor windows clone (complete with its own XML registry and breakages) this
is improvement.... New GCC that breaks 20% of packages - another
improvement.... NFS with added yummy locking that doesn't work ... better
still .... Meanwhile the linux hype machine is claiming the Linux desktop
is ready, well maybe ... if you only want to run gimp ?
Mark Johnson
> Im coming to the conclusion that the Linux desktop hype is a sack of
> shit.....
Really?
> Im running Fedora core 3/4 and Debian across 8 machines and its all
> starting to piss me off to the point im considering moving back to
> windows.
Mistake number 1, if you wan't stability and sanity don't mix distro's.
Move back to Windows if you like, do us a favour though run a mixture of
Windows 3.11, Windows 95 and Windows XP and come back and report your
findings.
I am sucessfully running 750 Debian desktops accross 340 locations,
there are few issues, its very stable and I only spend about 5 days a
month on average supporting them all. It *JUST WORKS*.
> Some examples:
>
> 1) Often when I yum update something breaks - often something difficult to
> fix
Can't vouce for Yum, apt-get works for me.
> 2) Updates causes firefox to lose its java plugin
Not here, I do the java plugin manually and make sure after an update
the library is copied into the right place.
> 3) Newer java runtime causes firefox not start, start from a console for
> NO debug output whatsoever - now thats well designed for "cutting edge
> browser"
>
> 4) Pan misses and duplicates characters as I type this.
Not had these issues
> 5) Open-office crashes and is not backward compatible with a lot of xls
> files.
Not sure about the crashes, have you reported them? Also Excel is
infamous for backward compatibility problems between its own versions,
Openoffice.org is not alone in this.
>
> 6) NFS and evolution interact - the "stateless server" NFS server causes
> fucking unhelpful error messages about lock ioctl or some such in
> evolution. Restarting the client doesnt fix it - you have to restart the
> server (or at least NFS service).
NFS is not the greatest service, try some others, Samba
> 7) Gnome now suffers window lockage just like windows. Open a dialog in
> one application blocks the opening of dialog in another.
Not had much of that either, though I admit it is not unknown. But you
can at least ssh in and kill gnome.
>
> 8) SSH setup isn't consistent across linux distros - some do X11
> forwarding others dont - some set the DISPLAY variable others dont ...
> very fucking annoying if you have lots of machines to admin.
Again distros' setup may vary.
>
> 9) The only thing constant about administering apache across distros is
> the name - every other fucking thing has a different path/filename...
See above
>
> 10) Evolution is the best visual client - but you still cant drag and drop
> attachments from within KDE .... desktop integration pha !
Yep, though there is work going on to fix this.
>
> Im I alone in having these day to day problems - im getting the impression
> that things are getting worse not better as the months tick by.
You are alone..
Try looking up cfengine2 and / or Fully Automated Installer
Decide on 1 distro, Debian works for us.
MarkJ
Phil Stovell
> Pan misses and duplicates characters as I type this.
That happens to me too. When I'm pissed.
--
Phil Stovell, South Hampshire, UK
Tony Davison
>
>
> On Mon, 26 Sep 2005 17:30:44 +0000, nope wrote:
>
>> Pan misses and duplicates characters as I type this.
>
> That happens to me too. When I'm pissed.
I lost the 'm' key the other day, cake crumb.
--
Big Tone
Tony van der Hoff
<pan.2005.09.27....@lycos.co.uk>
Which just goes to show how crappy linux is. Any OS worth its salt would be
able to recover from a cake-crumb error. ;-(
Phil Stovell
You obviously found it, toast crust.
Nix
On Tue, 27 Sep 2005, Unknown murmured woefully:
> On Tue, 27 Sep 2005 11:34:25 +0000, Neil Ellwood wrote:
>
>> On Mon, 26 Sep 2005 17:30:44 +0000, nope wrote:
>>
>>> Im coming to the conclusion that the Linux desktop hype is a sack of
>>> shit.....
I'm coming to the conclusion that you couldn't clearly report a problem
if your life depended on it.
>>> 1) Often when I yum update something breaks - often something difficult to
>>> fix
`Something'. More details please. We can't fix `something'.
>> Use google to find out what you are doing wrong.
> [[ Ahh I see ... running yum update makes ME wrong - not the fucking
> packagers - or linux itself - or the new GCC, nahhh ... its me ...
If any of the components you had named were totally broken, *thousands*
of people would have reported it (as we saw with the pointer aliasing
bugs in GCC-4.0.0, fixed in 4.0.1).
This has not happened.
Therefore the common factor and likely cause of the bug is you: your
setup, or something unusual you are doing that nobody else is doing.
(The latter is not unlikely: most of the bugs I've found have been
of that nature.)
> google to fix it, wow thats an idea i've never had - thank you for your
> words of wisdom.
An astonishing number of people have never thought of it.
>>> 2) Updates causes firefox to lose its java plugin
>> Has never happened to me (I have gone from man 7.2 right through to
>> Mandriva 2006 beta 2)
> Your not running either of two distros i've mentioned, this makes yours
> the comments worthless.
Most distributions use 99% similar software, including Firefox and
especially including the (binary-only) Java plugin.
His comments are not worthless: yours, on the contrary, are unnecessarily
combative. Do you *always* ask for help by attacking people?
>>> 3) Newer java runtime causes firefox not start, start from a console for
>>> NO debug output whatsoever - now thats well designed for "cutting edge
>>> browser"
>> If your not happy with that got to Epiphany.
strace -f output, please:
strace -o /tmp/blah -f /usr/bin/firefox
and send the last thousand lines or so of the strace.
P(0.999) this is a config problem, probably related to Firefox/Mozilla's
utterly odious `registry' nonsense. (Firefox and Moz are really very
un-Unixlike in many ways, as Pike as noted.)
>>> 4) Pan misses and duplicates characters as I type this.
>> What have You done to cause this - it has never happened to me.
> Who ho ...
Boy, that's constructive. Precise symptoms, please. Is every character
duplicated? Every other character missed? Is it totally random? Is the
CPU loaded while you do it?
>>> 6) NFS and evolution interact - the "stateless server" NFS server causes
>>> fucking unhelpful error messages about lock ioctl or some such in
>>> evolution. Restarting the client doesnt fix it - you have to restart the
>>> server (or at least NFS service).
>> As with the other items above you haven't supplied ANY info.
> [[I don't get ANY info - thats it ... a stupid error dialog picking up
> email, something like "unable to ioctl lock" or similar - NOTHING more, no
This error does not exist anywhere I can find, and there are dozens of
similar-sounding errors in this area.
We need at *least*:
- whether you are using the userspace or kernel-space NFS server
- your NFS version (NFSv2, NFSv3)
- the result of an `rpcinfo -p {clientname}' on the client and an
`rpcinfo -p {servername}' on the client
- the kernel version
- the email program you're using
- the exact error message
But it is much better to report this to your distribution vendor and/or
the linux-nfs mailing list: see <http://nfs.sourceforge.net/>, although
if you approach them with the attitude you've displayed here it is
vanishingly unlikely that you'll get any help.
> filenames, no paths, no context, no error logs...
Locking failures should be reported in the kernel logs, probably in
/var/log/kern*.
> Reboot the client and
> its in the same state, restart NFS on the server it works.
OK, so the server is losing track of client locks, probably.
> So NFS is
> stateless, except for the locking thats stateful
That is the protocol, yes. There have been bugs in this area: to my
knowledge as of Linux-2.6.12 none remain.
>>> 7) Gnome now suffers window lockage just like windows. Open a dialog in
>>> one application blocks the opening of dialog in another.
This is not correct. Your *window manager* may respect application
requests for modal dialogs: then again, it may not, and there's nothing
GNOME can do to override it. I'm not sure what Metacity, the (IMHO
horrible) GNOME window manager does by default. (I'm a fwvm2 and sawfish
user.)
>>> 8) SSH setup isn't consistent across linux distros - some do X11
>>> forwarding others dont - some set the DISPLAY variable others dont ...
>>> very fucking annoying if you have lots of machines to admin.
>> Don't get annoyed try Google to help yourself.
> [[I can help myself - im not complaining I don't know how, im complaining
> that stupid non-portable configurations make it time consuming and
> tedious.
I find SSH configs to be extremely portable between machines. The only
problem is when distros use radically different versions of SSH whose
support for features like environment variable exporting may not
even exist (e.g. OpenSSH 1.x). But this is rare: most distros stay
reasonably up-to-date.
>>> 10) Evolution is the best visual client - but you still cant drag and
>>> drop attachments from within KDE .... desktop integration pha !
>> Get programming skills going and do it.
> That has to be single most stupid thing any fuckwit has said so far. My
> programming skills will enable me to steer the compatibility of a KDE
> desktop and a gnome application will it? I steer the two projects and draw
> up the standards do I ?
The mailing lists are open; you are welcome to participate, although the
chances of anyone listening to you unless you grow a clue and some
diplomacy is minimal. There is no Super-S33kr1t Desktop Conspiracy.
>>> Im I alone in having these day to day problems - im getting the
>>> impression that things are getting worse not better as the months tick
>>> by.
>> YES
>
> Ahh silly me ... as KDE slides into bloat
Evidence?
> New GCC that breaks 20% of packages - another
> improvement....
I sometimes wonder why we bother writing the GCC release notes, which
say quite clearly when the C++ ABI changes and requires rebuilding/
relinking of all C++ applications (this does not happen every release,
but fairly frequently). I wonder if you forced RPM to ignore dependency
conflicts, or something... that's dangerous at the best of times, and
with something as core to the system as GCC, it is very stupid.
(If you didn't do that, then I don't have a clue what you did ---
because you never bothered to clearly describe it. Spotting a pattern
here? To do with your worse-than-useless bug reports? Yes, if you never
report a bug, and nobody else ever sees it, it will never be fixed. They
don't hand out psychic helmets with the Linux Developer's Cabal
membership badges anymore.)
> still .... Meanwhile the linux hype machine is claiming the Linux desktop
Claiming that anything to do with Linux is centrally-organised enough to
deserve the moniker `machine' is really quite laughable. You have no
clue how the community is organized at all, do you?
> is ready, well maybe ... if you only want to run gimp ?
GIMP is actually extremely complex with a very large number of
dependencies. So having it working is a good sign.
--
`One cannot, after all, be expected to read every single word
of a book whose author one wishes to insult.' --- Richard Dawkins
Unknown
> I'm coming to the conclusion that you couldn't clearly report a problem
> if your life depended on it.
problems - I was not after support, I was simply asking if I was the only
one to consider such fundamental problems a show stopper for this mythical
"ready" linux desktop.
>
>>>> 1) Often when I yum update something breaks - often something difficult to
>>>> fix
>
> `Something'. More details please. We can't fix `something'.
Over the past 8 months, gcc and X in that order.... The X package
overwrote my matrox driver with its own and lost dual head, gcc got out of
sync with one of its libs. Before you ask, no ... I ONLY used yum to
update it.
Yes, it is now fixed - im not struggling .... just swearing and wondering
if im pissing away more time than the results are worth.
>>>> 2) Updates causes firefox to lose its java plugin
>>> Has never happened to me (I have gone from man 7.2 right through to
>>> Mandriva 2006 beta 2)
>> Your not running either of two distros i've mentioned, this makes yours
>> the comments worthless.
>
> Most distributions use 99% similar software, including Firefox and
> especially including the (binary-only) Java plugin.
> His comments are not worthless: yours, on the contrary, are unnecessarily
> combative. Do you *always* ask for help by attacking people?
[[Im not asking for help - im pointing out that in my opinion the Linux
desktop is over hyped, I was simply seeing if anyone else felt the same.
So far almost everyone who responded has missed the point - in a very Unix
way. Im not complaining that I cant fix, im complaining that for a "ready
for business linux desktop" I should not have to KEEP fixing it !!
My servers are linux, and apart from the fact no two distros seem to have
the same defaults for anything .... apart from not being able to move a
web site from Redhat to Debian and back again at will (because the apache
modules don't share a common configuration etc etc etc) ... apart from all
the nagging day to day fixes and hand patching source to make things
build (even now), working out what security features have changed this
week ... apart from all of that they work.
However my office with 3 linux desktops seems to give me no end of grief -
and its not exactly cutting edge work. Basic stuff like networking doesn't
play well. Open office is more hype than reality, the desktop(s) are part
of the problems - two independent print managers that almost work, two
desktops with two APIs that almost sit together.
>
>>>> 3) Newer java runtime causes firefox not start, start from a console for
>>>> NO debug output whatsoever - now thats well designed for "cutting edge
>>>> browser"
>>> If your not happy with that got to Epiphany.
>
> strace -f output, please:
>
> strace -o /tmp/blah -f /usr/bin/firefox
>
> and send the last thousand lines or so of the strace.
>
I had already fixed this when I posted - I wasn't asking for fix, I was
just saying how unhelpful it is... I had to symlink the jre_1.5.0_04
plugin back to the jre_1.5.0_02 plugin, then it starts ....
> P(0.999) this is a config problem, probably related to Firefox/Mozilla's
> utterly odious `registry' nonsense. (Firefox and Moz are really very
> un-Unixlike in many ways, as Pike as noted.)
>
>>>> 6) NFS and evolution interact - the "stateless server" NFS server causes
>>>> fucking unhelpful error messages about lock ioctl or some such in
>>>> evolution. Restarting the client doesnt fix it - you have to restart the
>>>> server (or at least NFS service).
>>> As with the other items above you haven't supplied ANY info.
>> [[I don't get ANY info - thats it ... a stupid error dialog picking up
>> email, something like "unable to ioctl lock" or similar - NOTHING more, no
>
> This error does not exist anywhere I can find, and there are dozens of
> similar-sounding errors in this area.
Hence the words "or similar" ... I cant remember the exact line - I was
two busy swearing .... The thing that pissed me off more than locking not
working was that evolution (like so many others) didn't give me name of
the file that it was trying to lock. I *HATE* "I cant do that errors"
because it just spawns the question "cant do WHAT !"
>
> We need at *least*:
>
> - whether you are using the userspace or kernel-space NFS server - your
> NFS version (NFSv2, NFSv3)
> - the result of an `rpcinfo -p {clientname}' on the client and an
> `rpcinfo -p {servername}' on the client
> - the kernel version
> - the email program you're using
> - the exact error message
>
> But it is much better to report this to your distribution vendor and/or
> the linux-nfs mailing list: see <http://nfs.sourceforge.net/>, although
> if you approach them with the attitude you've displayed here it is
> vanishingly unlikely that you'll get any help.
>
>> filenames, no paths, no context, no error logs...
>
> Locking failures should be reported in the kernel logs, probably in
> /var/log/kern*.
You make my point well. The hype says linux on the desktop is ready, my
experience says its not. What office worker is possibly going to make it
one piece through this lot, how many day to day windows type admins could
submit a report in that kind of detail.......
> That is the protocol, yes. There have been bugs in this area: to my
> knowledge as of Linux-2.6.12 none remain.
[jon@jonspc ~]$ uname -a
Linux jonspc 2.6.12-1.1372_FC3 #1 Fri Jul 15 00:59:10 EDT 2005 i686 athlon i386 GNU/Linux
>>>> 7) Gnome now suffers window lockage just like windows. Open a dialog
>>>> in one application blocks the opening of dialog in another.
>
> This is not correct. Your *window manager* may respect application
> requests for modal dialogs: then again, it may not, and there's nothing
> GNOME can do to override it. I'm not sure what Metacity, the (IMHO
> horrible) GNOME window manager does by default. (I'm a fwvm2 and sawfish
> user.)
>
>>>> 8) SSH setup isn't consistent across linux distros - some do X11
>>>> forwarding others dont - some set the DISPLAY variable others dont
>>>> ... very fucking annoying if you have lots of machines to admin.
>>> Don't get annoyed try Google to help yourself.
>> [[I can help myself - im not complaining I don't know how, im
>> complaining that stupid non-portable configurations make it time
>> consuming and tedious.
>
> I find SSH configs to be extremely portable between machines. The only
> problem is when distros use radically different versions of SSH whose
> support for features like environment variable exporting may not even
> exist (e.g. OpenSSH 1.x). But this is rare: most distros stay reasonably
> up-to-date.
Why cant the distros just use a sensible set of defaults from the ssh
people ??? What the fuck is the point in having a portable machine
independent encrypted networking protocol, then making each distro not
"just work" with the next... what is that all about ? M/Soft must
be laughing their arses off at this mess, I would be laughing if I wasn't
swimming through this shit.... The Unix answer is always "JUST"
re-configure X,Y,Z.... by the time i've done this to 6 machines I now have
6 sets of configuration to backup, 6 sets of configuration to maintain, 6
sets of configuration to remember to keep in sync.
The "Use only one distro" call is no help, I have 6 machines I control -
I also help admin another 8. I don't get to chose whats installed
on others peoples servers, but I still have to work with it. Yet I can
pick a windows machine at random and I notice two things, one Its almost
always has a sensible configuration, two .. that configuration differs
very little from defaults.
>>>> 10) Evolution is the best visual client - but you still cant drag and
>>>> drop attachments from within KDE .... desktop integration pha !
>>> Get programming skills going and do it.
>> That has to be single most stupid thing any fuckwit has said so far. My
>> programming skills will enable me to steer the compatibility of a KDE
>> desktop and a gnome application will it? I steer the two projects and
>> draw up the standards do I ?
>
> The mailing lists are open; you are welcome to participate, although the
> chances of anyone listening to you unless you grow a clue and some
> diplomacy is minimal. There is no Super-S33kr1t Desktop Conspiracy.
>
>>>> Im I alone in having these day to day problems - im getting the
>>>> impression that things are getting worse not better as the months
>>>> tick by.
>>> YES
>>
>> Ahh silly me ... as KDE slides into bloat
>
> Evidence?
SPEED !
>
>> New GCC that breaks 20% of packages - another
>> improvement....
>
> I sometimes wonder why we bother writing the GCC release notes, which
> say quite clearly when the C++ ABI changes and requires rebuilding/
> relinking of all C++ applications (this does not happen every release,
> but fairly frequently). I wonder if you forced RPM to ignore dependency
> conflicts, or something... that's dangerous at the best of times, and
> with something as core to the system as GCC, it is very stupid.
Nope ! "rpm --install xyz.rpm" or just plain'ol "yum update"
Makes me wonder if Gentoo got the right ideas after all ? I've never
had time to try it.
I think part of the problem is the distros (apart from Debian) are
developed with the assumption that you have time or motivation to
re-format, re-install and copy back for every major point release.... they
let you "upgrade" - yet the upgraded machines never seem to be quite as
reliable as a fresh install.
> (If you didn't do that, then I don't have a clue what you did ---
> because you never bothered to clearly describe it. Spotting a pattern
> here? To do with your worse-than-useless bug reports? Yes, if you never
> report a bug, and nobody else ever sees it, it will never be fixed. They
> don't hand out psychic helmets with the Linux Developer's Cabal
> membership badges anymore.)
Im not making bug reports, I did submit bug reports for the stuff stat
mattered, and to be fair most of it was fixed.
My point was simply Linux desktop ready ?? I say no - you say ?? ....
and NFS stateless ?? Nahhh
Firefox fails with no debug information or errors - professional ??
Evolution complains about file locking - but which file ??
Its also possible that my problem is that somebody as over worked as
myself shouldn't have to maintain something as time consuming as Linux !
Vic
> for business linux desktop" I should not have to KEEP fixing it !!
That's true. The point that *you* seem to be missing is that you seem to
be in a very small minority in terms of the problems you're seeing.
> apart from not being able to move a
> web site from Redhat to Debian and back again at will
See, all my servers are RedHat (RH-derived, anyway). Except one, which
is Debian. And I don't see the same problem...
> apart from all
> the nagging day to day fixes and hand patching source to make things
> build (even now)
See, I don't do that sort of thing. I don't hand-patch code unless I
want to extend its functionality. And I don't have to problems you do...
> working out what security features have changed this
> week
I do keep an eye on that - and invariably, a fixed piece of code is
available to me to sort out any issues I might be facing. Even if I'm
not facing them.
> ... apart from all of that they work.
Had you considered that your problems could be down to your "hand
patching source to make things build (even now)"?
> Basic stuff like networking doesn't play well.
I've never had a system easier to work with on the networking front.
> Open office is more hype than reality
I find it difficult to "sell" to Windows-users because of the lack of
something Access-like. But the rest of it works as well as I need it to.
> The hype says linux on the desktop is ready, my
> experience says its not.
Your experience is decidedly atypical. Whether or not Linux is ready for
the desktop (and I'm inclined to say it's not - yet), what happens on
your systems is not run-of-the-mill; you might want to find a better
solution than blaming the software.
> > I find SSH configs to be extremely portable between machines. The only
> > problem is when distros use radically different versions of SSH whose
> > support for features like environment variable exporting may not even
> > exist (e.g. OpenSSH 1.x). But this is rare: most distros stay reasonably
> > up-to-date.
>
> Why cant the distros just use a sensible set of defaults from the ssh
> people ???
They do. See the point above - it's talking about using very old
versions. This is not a problem that occurs in reality.
> What the fuck is the point in having a portable machine
> independent encrypted networking protocol, then making each distro not
> "just work" with the next... what is that all about ?
Imagination?
All my boxes "just work" together just as soon as I open the port in the
firewall. That's the right way for it to happen.
> M/Soft must
> be laughing their arses off at this mess
There is no mess - outside of your network.
> Yet I can
> pick a windows machine at random and I notice ... Its almost
> always has a sensible configuration
Now I know this is a troll. Windows machines almost never have sensible
configurations - they have wide-open configurations. There's a big
difference.
> Nope ! "rpm --install xyz.rpm" or just plain'ol "yum update"
What about "hand patching source to make things build (even now)"?
> Its also possible that my problem is that somebody as over worked as
> myself shouldn't have to maintain something as time consuming as Linux !
That's probably correct. You spend loads of time maintaining it, and
have serious problems. I spend very little time maintaining my systems,
and have reliable, up-to-date servers. Are you seeing the same
correlation I am?
Vic.
--
Posted via Mailgate.ORG Server - http://www.Mailgate.ORG
Nix
> O
>> I'm coming to the conclusion that you couldn't clearly report a problem
>> if your life depended on it.
> [[I could, but this was written several hours *after* fixing most of these
> problems - I was not after support, I was simply asking if I was the only
> one to consider such fundamental problems a show stopper for this mythical
> "ready" linux desktop.
Since you didn't describe the nature of the problems, you've asked an
unanswerable question. In order to determine if the cause of a problem
is systematic or not, we have to know *what the problem is*.
>>>>> 1) Often when I yum update something breaks - often something difficult to
>>>>> fix
>>
>> `Something'. More details please. We can't fix `something'.
> Over the past 8 months, gcc and X in that order.... The X package
> overwrote my matrox driver with its own and lost dual head,
The question there becomes `why are you putting things in
/usr/X11R6/lib/X11/modules?'; that is the package manager's space, and
putting things into it without going via the package manager is risky.
> gcc got out of
> sync with one of its libs. Before you ask, no ... I ONLY used yum to
> update it.
Your distribution appears to have very broken dependencies, then. I'm
amazed that nobody reported this already.
> Yes, it is now fixed - im not struggling .... just swearing and wondering
> if im pissing away more time than the results are worth.
Well, on a Linux newsgroup people will tend to say `no'. Duh.
>> His comments are not worthless: yours, on the contrary, are unnecessarily
>> combative. Do you *always* ask for help by attacking people?
> [[Im not asking for help - im pointing out that in my opinion the Linux
> desktop is over hyped, I was simply seeing if anyone else felt the same.
Well, I've been using a `Linux desktop' without trouble for eight years ---
but I'm a techie. I'd so far not recommend most Linux distros to non-techies
without a techie nearby --- but I'd disrecommend Windows even more intensely.
MacOS X seems to be the best recommendation for `just make it work dammit'
users right now.
Xandros is not half bad. Ubuntu is getting there but still has rough
edges. Mandrake is fairly good but has a high emphasis on flashiness
still, at times at the expense of stability. (That's my impression,
anyway.)
(But personally I don't use any of these except in chroots anyway.)
> So far almost everyone who responded has missed the point - in a very Unix
> way. Im not complaining that I cant fix, im complaining that for a "ready
> for business linux desktop" I should not have to KEEP fixing it !!
Agreed. It looks like you fubared your system in a rather fundamental fashion
early on, probably breaking yum and/or the package manager's dependency
database, and from then on every package you installed would be risking
clashes with what was *really* on the system. (I've done this myself via
an accidental rm on an RH9 box, and, yes, recovery is nasty. Non-techies
should probably preserve /home and reinstall. But then that's what they'd
be likely to do anyway...)
> My servers are linux, and apart from the fact no two distros seem to have
> the same defaults for anything ....
Well, no. If they used the same defaults for everything, they'd be pretty
much the same distro. It's called `choice'.
Newbie desktop users aren't likely to be running networks with dozens of
machines with different distros on them. Neither are most large-scale
sysadmins: they'll be imaging the same distro onto heaps of nearly-
identical boxes.
It's only the midscale experimentalist who has trouble --- and, well,
he'll keep at it no matter what happens, since learning things is the
whole point for that sort of loony. (Like, er, me.)
> hand patching source to make things
> build (even now),
This was probably (mostly) due to your GCC upgrade. A significant amount
of C++ stuff still needs patching to build with GCC 4.0; more patches
will be needed for 4.1. Blame the authors for not knowing every detail
of the C++ Standard, or GCC for providing a C++ compiler before every
detail of standards-compliance was sorted out. (OK, so perhaps that
isn't so terribly blameworthy, not least because if we did that G++
would still not have been released... nearly twenty years after its
actual initial release date.)
> working out what security features have changed this
> week
Until Shapiro's capability-based stuff takes off, the patch treadmill is
unavoidable for users of *all* OSes. The difference is that the Linux world
at least provides patches soon after the breakage.
> However my office with 3 linux desktops seems to give me no end of grief -
> and its not exactly cutting edge work. Basic stuff like networking doesn't
> play well.
Networking, in particular NFS, can cause interesting interlocking
failures and deadlocks if you have a tightly-interconnected network. What I
did was to try spontaneously turning single machines off, find out what
unexpectedly broke on the other machines, and fix it. (Things like $PATH
pointing through a directory which was pointed to via a symlink underneath
an NFS-mounted dir, some proxy-arp timeout problems, et al.)
Another important test is to flip the house/office power off and on and
be sure that everything restarts. Crap PSUs that stay off until manually
restarted are a problem: the converse is also a problem because rapid
successive power cuts can cause major hardware damage (we really need
something which holds the machine off for a minute or so after a power
cut and then turns it on). Machines NFS-mounting things from each other
are interesting here: the `bg' option can be useful. There are other
classic deadlock situations: a good first attempt is to start local
stuff first, then network servers, and *only then* stuff that depends on
any remote services. It's a classic lock-hierarchy problem, and init
systems (like init-ng) which allow the explicit modelling of that lock
hierarchy are finally starting to turn up.
Unfortunately modelling this yourself is annoyingly difficult unless
you have some experience in concurrent systems design :( but, hey,
it's a good excuse to learn about it!
> Open office is more hype than reality,
Nah, it exists, therefore it is not hype.
(Now *bloat* I'll buy. Ye gods it's fat. All I use it for is to read the
odd Word document that mswordview and antiword and friends can't
decrypt...)
> the desktop(s) are part
> of the problems - two independent print managers that almost work, two
What? GNOME and KDE don't have `print managers'. They have fairly thin
interfaces to CUPS / LPRng / BSD lpr. Stick a job in from one and it's
visible from the other (generally).
CUPS, LPRng and lpr all have different advantages, of course: CUPS gives
you nice control over printer capabilities and point-and-clickydom,
LPRng gives you insane customizability via /etc/printcap's rebarbative
user interface and astonishingly powerful filtering and routing
capabilities (I use mine to do things totally unrelated to printing like
doing background batch raytraces, and I've seen people using it to
distribute compilation jobs over a compile farm!), and BSD lpr, er, lpr
is, um, traditional. ;}
> desktops with two APIs that almost sit together.
Well, the APIs are different and the standards are converging. There is
lots of cooperation, and the X people are far more responsive than they
used to be, so things are improving.
> I had already fixed this when I posted - I wasn't asking for fix, I was
> just saying how unhelpful it is... I had to symlink the jre_1.5.0_04
> plugin back to the jre_1.5.0_02 plugin, then it starts ....
Well, yes, that's going to break things. If you have broken links to
random shared libraries, things are going to break.
I've never seen this much package-related breakage on a single system
at once unless I'd done something silly (of course I've done that before;
accidental rm's of /usr/bin tend to do bad things, but I didn't blame the
*system*. On the contrary it was amazing how well the system worked while
I recovered things. Even my X sessions kept humming along.)
>>> [[I don't get ANY info - thats it ... a stupid error dialog picking up
>>> email, something like "unable to ioctl lock" or similar - NOTHING more, no
>>
>> This error does not exist anywhere I can find, and there are dozens of
>> similar-sounding errors in this area.
> Hence the words "or similar" ... I cant remember the exact line - I was
Well, if you don't note the error *nobody can fix it*, or even
commiserate properly (I've seen similar errors, but only some time in
the past, and I don't know if they were actually similar, so mentioning
them is pointless).
> working was that evolution (like so many others) didn't give me name of
> the file that it was trying to lock. I *HATE* "I cant do that errors"
> because it just spawns the question "cant do WHAT !"
Well, my opinion of Evolution is unprintable. (I don't think much of
GNOME as a whole --- its user-interface philosophy is the diametric
opposite of mine --- but Evolution, ick.)
>> Locking failures should be reported in the kernel logs, probably in
>> /var/log/kern*.
> You make my point well. The hype says linux on the desktop is ready, my
> experience says its not. What office worker is possibly going to make it
> one piece through this lot, how many day to day windows type admins could
> submit a report in that kind of detail.......
That's why distribution vendors like RH ship systems which come with
support contracts that have *them* acquiring this sort of data for you,
or telling you exactly how to do it.
i.e., if you pay for a support contract, you'll get it. Otherwise, *you*
have to do the data collection.
>> That is the protocol, yes. There have been bugs in this area: to my
>> knowledge as of Linux-2.6.12 none remain.
>
> [jon@jonspc ~]$ uname -a
> Linux jonspc 2.6.12-1.1372_FC3 #1 Fri Jul 15 00:59:10 EDT 2005 i686 athlon i386 GNU/Linux
Interesting. Of course this is about 5% of the info I asked for, and
without the rest I can't draw many conclusions (lockd and statd not
running on client *and* server could cause symptoms involving locking
problems, but without knowing more about the error I can't narrow down
the possible cause more than that).
>> I find SSH configs to be extremely portable between machines. The only
>> problem is when distros use radically different versions of SSH whose
>> support for features like environment variable exporting may not even
>> exist (e.g. OpenSSH 1.x). But this is rare: most distros stay reasonably
>> up-to-date.
>
> Why cant the distros just use a sensible set of defaults from the ssh
> people ???
Newsflash: *they do*.
But the SSH defaults are, ahem, insanely paranoid for many applications.
No X forwarding, no agent forwarding, discarding of all env vars: some
distros agree with this policy, many do not. (For myself, I strip
DISPLAY when sshing to some untrusted hosts, and not otherwise.
There's no way a distro vendor could know which hosts I trust!)
> What the fuck is the point in having a portable machine
> independent encrypted networking protocol, then making each distro not
> "just work" with the next...
They do. You can ssh from one box to another pretty much fine with any
of the distro configs. If you care about X connection forwarding and
that sort of thing, well, that's a site policy matter, really, as X
forwarding can open you up to certain types of attack.
> The Unix answer is always "JUST"
> re-configure X,Y,Z.... by the time i've done this to 6 machines I now have
> 6 sets of configuration to backup, 6 sets of configuration to maintain, 6
> sets of configuration to remember to keep in sync.
Learn about cfengine, rsync, hell, even cp or NFS-mounting or centrally
version-controlling /etc. There are *countless* ways to eliminate this
burden. The reason why nobody else running a substantial network
complains is because everyone else is using the tools that exist rather
than just moaning.
> The "Use only one distro" call is no help, I have 6 machines I control -
> I also help admin another 8. I don't get to chose whats installed
> on others peoples servers, but I still have to work with it. Yet I can
> pick a windows machine at random and I notice two things, one Its almost
> always has a sensible configuration, two .. that configuration differs
> very little from defaults.
Are you *seriously* saying that any two randomly selected Windows
systems will have even slightly similar configurations? *bwahahaaaaa*
the only way to keep the admin load sane on Windows boxes is to almost
totally ban installing or reconfiguring software on it and to mandate
exactly the same Windows version across the organization. (This is
generally impractical.)
>>> Ahh silly me ... as KDE slides into bloat
>>
>> Evidence?
> SPEED !
Startup speed? Execution speed?
(If the former, use prelink, dammit, KDE was why it was implemented in
the first place. Most distros these days do that by default. If the
latter, profile, dammit: I've seen no substantive speed problems, even
on a P233.)
> I think part of the problem is the distros (apart from Debian) are
> developed with the assumption that you have time or motivation to
> re-format, re-install and copy back for every major point release.... they
> let you "upgrade" - yet the upgraded machines never seem to be quite as
> reliable as a fresh install.
Yes. I dislike this a lot as well. (This is why I use Gentoo, Debian, or
home-rolled systems for anything I expect to have to massively
hack. Generally such are experimental machines anyway.)
I guess you can mostly blame RPM for this, but not entirely.
> My point was simply Linux desktop ready ?? I say no - you say ?? ....
It depends what you mean by `desktop ready'.
It's ready for random-user use on networks with sysadmins who do the
sysadminish stuff. In fact it's far preferable to Windows, because users
can't shoot themselves in the foot in so many ways.
It's not ready for your grandma's home machine. (But nobody sane has
ever claimed otherwise.)
> and NFS stateless ?? Nahhh
Well, no. NFS ain't called the Nightmare File System for nothing.
People have been getting halfway in writing better distributed FSen
for ages, but either they die young (e.g. Intermezzo) or they're very
non-Unixlike (Samba, AFS, Coda), or they're very non-scalable (Coda).
I'll have to try Lustre one of these days. Perhaps my wait is over...
> Firefox fails with no debug information or errors - professional ??
There is an envrionment variable you can set to get more verbosity. It
used to be way too verbose; I guess it's gone too far the other way.
And no, it's not `professsional'. I've used `professional' systems:
they're *far* more annoying. Had this been a `professional' system you'd
have to do something like constructing a massive XML file yourself and
sticking it in a poorly-documented location to describe the exact
location and function of all of your plugins, and then watch it get
overwritten whenever the system felt like it. There would be expensive
positions advertised for `Firefox Administrators' and so on. (This is
apparently a good thing because it creates a market in Firefox
Administrators. See, e.g., the antivirus industry.)
> Its also possible that my problem is that somebody as over worked as
> myself shouldn't have to maintain something as time consuming as Linux !
You need to learn how to use the automation tools (like cfengine, et al)
that you already have.
Roger Leigh
> Well, my opinion of Evolution is unprintable. (I don't think much of
> GNOME as a whole --- its user-interface philosophy is the diametric
> opposite of mine --- but Evolution, ick.)
RESOLVED WONTFIX
(Sorry, couldn't resist; I'm pretty fed up with GNOME development at
the moment. It's as stagnant as my garden pond.)
Nix
> Nix <nix-ra...@esperi.org.uk> writes:
>
>> Well, my opinion of Evolution is unprintable. (I don't think much of
>> GNOME as a whole --- its user-interface philosophy is the diametric
>> opposite of mine --- but Evolution, ick.)
>
> RESOLVED WONTFIX
ROFL!
> (Sorry, couldn't resist; I'm pretty fed up with GNOME development at
> the moment. It's as stagnant as my garden pond.)
Has the developer momentum swept back to KDE again? I know GNOME was all
gung-ho in the 1.3.x and pre-2.0 days: has it quagmired?
--
`Next: FEMA neglects to take into account the possibility of
fire in Old Balsawood Town (currently in its fifth year of drought
and home of the General Grant Home for Compulsive Arsonists).'
--- James Nicoll
Nick Kew
> Has the developer momentum swept back to KDE again? I know GNOME was all
> gung-ho in the 1.3.x and pre-2.0 days: has it quagmired?
>
Well, if you take the projects that participated in the google
summer-of-code as a measure, KDE is thriving.
(we @apache got some good stuff out of that, too:-)
--
Not me guv
Roger Leigh
Hash: SHA1
Nix <nix-ra...@esperi.org.uk> writes:
> On Thu, 29 Sep 2005, Roger Leigh suggested tentatively:
>> Nix <nix-ra...@esperi.org.uk> writes:
>>
>>> Well, my opinion of Evolution is unprintable. (I don't think much of
>>> GNOME as a whole --- its user-interface philosophy is the diametric
>>> opposite of mine --- but Evolution, ick.)
>>
>> RESOLVED WONTFIX
>
> ROFL!
>
>> (Sorry, couldn't resist; I'm pretty fed up with GNOME development at
>> the moment. It's as stagnant as my garden pond.)
>
> Has the developer momentum swept back to KDE again? I know GNOME was all
> gung-ho in the 1.3.x and pre-2.0 days: has it quagmired?
Well, from my point of view, there are some fundamental design flaws
that they are completely unwilling to fix. Fixing them would break
the current ABI, but rather than having a development branch or some
place to fix them, the bug reports and patches are just dropped on the
floor. Unfortunately for me, without the fixes some things are simply
impossible to achieve.
It seems that RedHat (who employ nearly all the official developers)
are so busy trumpeting the ABI stability for "ISVs" they completely
forgot about fixing the real bugs, in order to make it really good and
useful for free software developers. I'm personally just about ready
to rewrite all my code in C++, despite the extreme "yuck" factor (I'm
not a C++ fan...).
The major bug is the split between glib and gobject. Quite a few of
the glib types should be full-fledged objects, or else their utility
is extremely limited. For example, GMain should be an object, and
GSource should be a GInterface, so that any object implementing it can
be used as an event source in the mainloop. Currently only structures
containing a GSource as their first member can be event sources, and
these are not objects. So many GLib types re-invent reference
counting and destruction notification it's untrue; being an object
would provide a unified interface and remove a whole lot of
wheel-reinvention.
I'd also like to see the basic container types implement a generic
GContainer interface, and implement proper iterators. Unless I
implement it myself, this will never happen. I just need more time in
each day... Check how many g_foo_foreach functions exist. A single
g_container_foreach could replace the lot.
Here's a small selection of nasties:
GObject is not const-correct, and they basically don't care:
http://bugzilla.gnome.org/show_bug.cgi?id=303543
This makes it impossible to write const-correct code /and/ enable lots
of GCC warnings without being buried in warning spew.
Some of the maintainers don't care about portability, even when the
current code is completely broken and dangerous:
http://bugzilla.gnome.org/show_bug.cgi?id=314067
(this one should get fixed, but when...)
Symmetric APIs, to allow code reuse is not desirable, even when a
4-line macro or function would suffice:
http://bugzilla.gnome.org/show_bug.cgi?id=317115
And some fixes are trivial, yet are never applied, leading to all
users writing stupid glue code:
http://bugzilla.gnome.org/show_bug.cgi?id=316961
http://bugzilla.gnome.org/show_bug.cgi?id=300610
(without this, you can't propagate GErrors back through closures; this
typically means throwing exceptions through signal handlers, something
that can be very useful: signal handlers do sometimes fail! Both the
GStreamer folks and myself need this, yet we both have to re-implement
it over and over--I have a copy in each gobject-using package I
maintain!)
Some core parts are basically unmaintained for years:
http://bugzilla.gnome.org/show_bug.cgi?id=166707
http://bugzilla.gnome.org/show_bug.cgi?id=308769
And even documentation improvements are rejected:
http://bugzilla.gnome.org/show_bug.cgi?id=310174
(the RESOLVED WONTFIX gets pretty disheartening after a while--that's
6 hours/2 evenings wasted in this case. If it didn't solve a problem
that was important to me, /why/ would I have bothered writing it?!?)
Sorry for that, I'm just a little bit riled about this. Submitting
patches is just like running against a brick wall... It's not like
I'm not prepared to put in the effort to fix things, but rather that
after spending the effort on investigating and writing a good patch,
it's summarily binned for no good reason.
Regards,
Roger
- --
Roger Leigh
Printing on GNU/Linux? http://gimp-print.sourceforge.net/
Debian GNU/Linux http://www.debian.org/
GPG Public Key: 0x25BFB848. Please sign and encrypt your mail.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.2 (GNU/Linux)
Comment: Processed by Mailcrypt 3.5.8+ <http://mailcrypt.sourceforge.net/>
iD8DBQFDPpqUVcFcaSW/uEgRAmUYAKC3Nm0EtagG+d09cCrRx/jAhXSixwCg6QLF
sNpW0pRymEt2mrW0/YKY6GI=
=8UlY
-----END PGP SIGNATURE-----
Nix
> Nix <nix-ra...@esperi.org.uk> writes:
>> Has the developer momentum swept back to KDE again? I know GNOME was all
>> gung-ho in the 1.3.x and pre-2.0 days: has it quagmired?
>
> Well, from my point of view, there are some fundamental design flaws
> that they are completely unwilling to fix. Fixing them would break
> the current ABI, but rather than having a development branch or some
> place to fix them, the bug reports and patches are just dropped on the
> floor. Unfortunately for me, without the fixes some things are simply
> impossible to achieve.
Well, there are some definite uglinesses in KDE too, which its equally
strict ABI compatibility rules forbid fixing. The difference is that
KDE breaks compatibility a bit more often (generally because Qt has
done so, so there's no disadvantage to KDE doing so).
But none are is ugly as libgnome or the thankfully now dead rev-of-the-
week libgal.
> It seems that RedHat (who employ nearly all the official developers)
> are so busy trumpeting the ABI stability for "ISVs" they completely
> forgot about fixing the real bugs,
I think we're just seeing typical we-don't-dare-release-stably-yetitis
(as seen in many many projects from GCC to GNU younameit to the Linux
kernel), only applied in a slightly different sphere. Instead of
saying `we don't dare call this stable yet', they're saying `we don't
have enough stuff built up to warrant an ABI break yet', past all
sanity.
> in order to make it really good and
> useful for free software developers. I'm personally just about ready
> to rewrite all my code in C++, despite the extreme "yuck" factor (I'm
> not a C++ fan...).
KDE's C wrappers are... embryonic. Eventually, hopefully there will be
some (I think the current plan is to build them off the `smoke' Qt
scripting bindings).
Myself, I like C++ to some extent (it's an excessively baroque language,
but nice compared to e.g. Java pre-1.5) but find what Qt has done to it
quite disgusting (`moc' in particular; hint, guys, this horrid
preprocessing step is *provably unnecessary*, see e.g. libsigc++ for a
cleaner implementation). However, it's tolerably disgusting because it
doesn't damage the *language* per se and the preprocessor is nowhere
near as vile as e.g. Oracle's Pro/C++ preprocessor.
> The major bug is the split between glib and gobject.
This was trumpeted as a major advance when it took place. A shame they
put the split in the wrong place (as you point out).
> Quite a few of
> the glib types should be full-fledged objects, or else their utility
> is extremely limited. For example, GMain should be an object, and
> GSource should be a GInterface, so that any object implementing it can
> be used as an event source in the mainloop.
It *isn't*? *boggle*
> Currently only structures
> containing a GSource as their first member
So they're actually *relying* on the interconvertibility of a pointer to
a structure and a pointer to its first member? Ye *gods*. And I thought
the GCC hackers were overreacting when they withdrew part of an
optimization because it would break this. No. Iccck.
Forget C++: C is truly disgusting. Combine this grot with things like
`arrays of N elements can be safely overflowed when they are the last
element in a structure', which renders bounds-checking almost impossible
in many cases, and, well, *shudder*.
> can be event sources, and
> these are not objects. So many GLib types re-invent reference
> counting and destruction notification it's untrue; being an object
> would provide a unified interface and remove a whole lot of
> wheel-reinvention.
Now if they were using C++ they could just use a shared_ptr<>... ;}
(that's my *other* gripe with Qt: its reinvention of scads of the C++
standard library. Thankfully the tide is reversing --- most of it was
only there because of the appalling quality of old standard C++
implementations --- and a lot of those barmy types like QList have been
eliminated in Qt 4, or turned into thin wrappers around the underlying
standard C++ types.)
> I'd also like to see the basic container types implement a generic
> GContainer interface, and implement proper iterators. Unless I
> implement it myself, this will never happen.
You'd think the Glib hackers would have noticed the absence of something
like this... We need an STL for C (but alas this is effectively
impossible).
> Here's a small selection of nasties:
>
> GObject is not const-correct, and they basically don't care:
> http://bugzilla.gnome.org/show_bug.cgi?id=303543
> This makes it impossible to write const-correct code /and/ enable lots
> of GCC warnings without being buried in warning spew.
They call const-correctness `clutter'?
Sheesh. Someone tell them about flash-memory embedded systems. (Also
about the extra optimizations we can do given const-correctness and
restrict-correctness, some quite significant. Things like automatically
locating functions that could be marked with the pure or const
attributes, and marking them. That lets you do things like common
subexpression elimination with *whole function calls*).
You are wrong in one part of that bug report, btw; constness cannot
always be safely cast away. It can only be cast away if the *original
declaration* of the pointed-to object in question was not const,
otherwise you're invoking undefined behaviour with no diagnostic
required. Of course there's no way bar convention for the called
function to tell how the pointed-to object was originally declared. It
can be very hard to tell if that's happening until the const object
lands in read-only memory and you get a crash, or, worse, a silent
non-updating of the object, as happens on some embedded platforms when
you try to update ROM.
As for `only pedants compile with warnings enabled', well, sheesh.
Which idiot said that? (The bad-mouthing of -pedantic in the GCC manual
vanished ages ago, and GCC itself is now compiled with -pedantic
-Werror.)
In any case, shouldn't good programmers *be* pedants when writing code?
It's not as if the computer can avoid being pedantic itself.
> Some of the maintainers don't care about portability, even when the
> current code is completely broken and dangerous:
> http://bugzilla.gnome.org/show_bug.cgi?id=314067
> (this one should get fixed, but when...)
Note that your fix is the *only* safe one for another reason: time_t is
required to an arithmetic type, but it need not be any of the
conventional types: it can expand to a weird compiler-internal type
that's only used for time_t. (The same is true of some macros,
e.g. NULL, where GCC actually does play such tricks, and regarding
fpos_t, the library has even more freedom. Need I say that GNOME broke
when glibc/libio moved to making fpos_t a structure? They really
*don't* learn...)
> Symmetric APIs, to allow code reuse is not desirable, even when a
> 4-line macro or function would suffice:
> http://bugzilla.gnome.org/show_bug.cgi?id=317115
You're *surprised* that the GNOME people, who speak approvingly of the
appalling Win32 API, don't understand cleanliness or elegance?
> And some fixes are trivial, yet are never applied, leading to all
> users writing stupid glue code:
> http://bugzilla.gnome.org/show_bug.cgi?id=316961
> http://bugzilla.gnome.org/show_bug.cgi?id=300610
What good reasoning. They resist it because they've always resisted it.
No actual logic or rationale provided.
> (without this, you can't propagate GErrors back through closures; this
> typically means throwing exceptions through signal handlers, something
> that can be very useful: signal handlers do sometimes fail! Both the
> GStreamer folks and myself need this, yet we both have to re-implement
> it over and over--I have a copy in each gobject-using package I
> maintain!)
I start to see why libgal and libgnome turned up :(
> Some core parts are basically unmaintained for years:
> http://bugzilla.gnome.org/show_bug.cgi?id=166707
Probably because the canvas's more advanced features were pretty much
only used by the GIMP, and the GIMP ended up forking its own version
anyway. So now there are no maintainers...
> http://bugzilla.gnome.org/show_bug.cgi?id=308769
Well, that `small fix' would require either replacing glib-gettextize
with a do-nothing wrapper, or changing *everyone's* autogen.sh. But
that's only a gnome-common change. More stick-in-the-mudism.
> And even documentation improvements are rejected:
> http://bugzilla.gnome.org/show_bug.cgi?id=310174
Sheesh! They *really* have a bad case of NIH syndrome, don't they?
(Either that or a variant on RMS's `leave internal functions undocumented'
thing from GNU Emacs: but at least *that* had a rationale before etc/DOC
was loaded lazily.)
> (the RESOLVED WONTFIX gets pretty disheartening after a while--that's
> 6 hours/2 evenings wasted in this case. If it didn't solve a problem
> that was important to me, /why/ would I have bothered writing it?!?)
Because the magic telepathy hat that all core GNOME developers get
issued with tells them that Other Developers / The Users Won't Want It.
(Or was that a magic arrogance hat?)
> Sorry for that, I'm just a little bit riled about this. Submitting
> patches is just like running against a brick wall... It's not like
> I'm not prepared to put in the effort to fix things, but rather that
> after spending the effort on investigating and writing a good patch,
> it's summarily binned for no good reason.
I expect the *real* reason is `I didn't write this'. :/
Roger Leigh
Hash: SHA1
Nix <nix-ra...@esperi.org.uk> writes:
> On Sat, 01 Oct 2005, Roger Leigh wrote:
>> Nix <nix-ra...@esperi.org.uk> writes:
>> It seems that RedHat (who employ nearly all the official developers)
>> are so busy trumpeting the ABI stability for "ISVs" they completely
>> forgot about fixing the real bugs,
>
> I think we're just seeing typical we-don't-dare-release-stably-yetitis
> (as seen in many many projects from GCC to GNU younameit to the Linux
> kernel), only applied in a slightly different sphere. Instead of
> saying `we don't dare call this stable yet', they're saying `we don't
> have enough stuff built up to warrant an ABI break yet', past all
> sanity.
Glib 2.0.0 has been around for over 3½ years now, and the API frozen
for much longer than that, so IMHO it's well past the time to break
it.
> Myself, I like C++ to some extent (it's an excessively baroque language,
> but nice compared to e.g. Java pre-1.5) but find what Qt has done to it
> quite disgusting (`moc' in particular; hint, guys, this horrid
> preprocessing step is *provably unnecessary*, see e.g. libsigc++ for a
> cleaner implementation).
I really do like libsigc++. But I don't like the lack of unity
between all the different C++ libraries. That's one thing GObject
gave you: properties, signals, object instantiation, reference
counting and two-phase object destruction (to break cycles). The STL
goes some way to compensate, but IMO templates are often the wrong
answer, since you polymorphism (RTTI etc.) out of the window, reducing
the flexibility of what can be done at runtime.
>> For example, GMain should be an object, and GSource should be a
>> GInterface, so that any object implementing it can be used as an
>> event source in the mainloop.
>
> It *isn't*? *boggle*
Quite.
>> Currently only structures containing a GSource as their first
>> member
>
> So they're actually *relying* on the interconvertibility of a pointer to
> a structure and a pointer to its first member? Ye *gods*. And I thought
> the GCC hackers were overreacting when they withdrew part of an
> optimization because it would break this. No. Iccck.
That assumption is made throughout GLib and GObject. In fact, it's
fundamental for the RTTI and single inheritance, which relies on this
for dynamic casting and polymorphism. The basic object "foo" looks
like this (+ == contains as first member):
Foo
+ GObject
+ GTypeInstance
+ GTypeClass
FooClass
+ GObjectClass
+ GTypeClass
+ GType
So this is pretty much directly equivalent to the C++ vtable and
typeid, so you can go from object->class->type
(object->vtable->typeid).
For for any object, *((GType *)(*(GTypeClass *)obj)) will give you the
type id, assuming contained structs will have the same offset 0 as the
containing struct/union.
What I really do like taking advantage of is the ability to be able to
instantiate arbitrary objects with nothing but a GType typeid and a
construction property list; no knowledge of the type size or structure
is required. In this respect, moving to C++ would be a step backward;
I'll probably look into Objective C first.
>> I'd also like to see the basic container types implement a generic
>> GContainer interface, and implement proper iterators. Unless I
>> implement it myself, this will never happen.
>
> You'd think the Glib hackers would have noticed the absence of something
> like this... We need an STL for C (but alas this is effectively
> impossible).
GLib provides GValue, a mechanism for storing and copying any value,
from fundamental C types to boxed types (C structs) and other GTypes
derived from GObject. If a container stored GValues, it could be
completely generic, at the expense of some performance (due to type
checking).
>> GObject is not const-correct, and they basically don't care:
[...]
> They call const-correctness `clutter'?
[...]
> You are wrong in one part of that bug report, btw; constness cannot
> always be safely cast away. It can only be cast away if the *original
> declaration* of the pointed-to object in question was not const,
> otherwise you're invoking undefined behaviour with no diagnostic
> required.
This is an implementation detail of GObject. All GObject-derived
objects must be constructed with g_object_new, and are dynamically
allocated with g_new(). As a result, they will always be writable
(unless you play games with mprotect).
(This is another gripe I have: there's no in-place construction á la
the STL. This would complicate the reference counting, but would give
much better control over memory allocation in e.g. containers.)
> In any case, shouldn't good programmers *be* pedants when writing code?
> It's not as if the computer can avoid being pedantic itself.
I thought being a pedant came with the territory...
>> Sorry for that, I'm just a little bit riled about this. Submitting
>> patches is just like running against a brick wall... It's not like
>> I'm not prepared to put in the effort to fix things, but rather that
>> after spending the effort on investigating and writing a good patch,
>> it's summarily binned for no good reason.
>
> I expect the *real* reason is `I didn't write this'. :/
:-/
Regards,
Roger
- --
Roger Leigh
Printing on GNU/Linux? http://gimp-print.sourceforge.net/
Debian GNU/Linux http://www.debian.org/
GPG Public Key: 0x25BFB848. Please sign and encrypt your mail.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.2 (GNU/Linux)
Comment: Processed by Mailcrypt 3.5.8+ <http://mailcrypt.sourceforge.net/>
iD8DBQFDSXoGVcFcaSW/uEgRAmmnAKDRU6LsiWiGw499KY78LmmHA2uA9QCg1Ke0
dboB5lO9+6eNj9Xt66rM9bA=
=h+Hi
-----END PGP SIGNATURE-----
Tony Houghton
Roger Leigh <${rleigh}@invalid.whinlatter.ukfsn.org.invalid> wrote:
> What I really do like taking advantage of is the ability to be able to
> instantiate arbitrary objects with nothing but a GType typeid and a
> construction property list; no knowledge of the type size or structure
> is required. In this respect, moving to C++ would be a step backward;
> I'll probably look into Objective C first.
Python's really nice if you don't mind using an interpreted language.
It's high level, so I find it a lot quicker and less fiddly than C (or
C++, but I generally avoid that now), but it has the feel of a "real
programming language", whereas perl feels more like a scripting
language.
--
The address in the Reply-To is genuine and should not be edited.
See <http://www.realh.co.uk/contact.html> for more reliable contact addresses.
Ewan Mac Mahon
> In <8764s64...@hardknott.home.whinlatter.ukfsn.org>,
> Roger Leigh <${rleigh}@invalid.whinlatter.ukfsn.org.invalid> wrote:
>
> Python's really nice if you don't mind using an interpreted language.
> It's high level, so I find it a lot quicker and less fiddly than C (or
> C++, but I generally avoid that now), but it has the feel of a "real
> programming language", whereas perl feels more like a scripting
> language.
>
need the speed, control or flexibility for you can just write that bit in C
and the rest in Python.
That said, Python's standard library is pretty good, so it doesn't happen
all that often.
Ewan
reclusive monkey
>>> That happens to me too. When I'm pissed.
>> I lost the 'm' key the other day, cake crumb.
>You obviously found it, toast crust.
Small buds of skunk for me... which explains a lot.
Geoffrey Clements
news:slrndkja08.klb.th...@realh.co.uk...
> Python's really nice if you don't mind using an interpreted language.
Python is really nice even if you *do* mind an interpreted language :-)
> It's high level, so I find it a lot quicker and less fiddly than C (or
> C++, but I generally avoid that now), but it has the feel of a "real
> programming language", whereas perl feels more like a scripting
> language.
>
Let the wars begin ...
--
Geoff
Roger Leigh
Hash: SHA1
Tony Houghton <this.addre...@realh.co.uk> writes:
> In <8764s64...@hardknott.home.whinlatter.ukfsn.org>,
> Roger Leigh <${rleigh}@invalid.whinlatter.ukfsn.org.invalid> wrote:
>
>> What I really do like taking advantage of is the ability to be able to
>> instantiate arbitrary objects with nothing but a GType typeid and a
>> construction property list; no knowledge of the type size or structure
>> is required. In this respect, moving to C++ would be a step backward;
>> I'll probably look into Objective C first.
>
> Python's really nice if you don't mind using an interpreted language.
> It's high level, so I find it a lot quicker and less fiddly than C (or
> C++, but I generally avoid that now), but it has the feel of a "real
> programming language", whereas perl feels more like a scripting
> language.
I liked Python, for the small amount of time I spent with it. I just
haven't fully grokked the function pass by sort-of-reference
semantics.
Unfortunately I need to be setuid, so I need a compiled language.
Regards,
Roger
- --
Roger Leigh
Printing on GNU/Linux? http://gimp-print.sourceforge.net/
Debian GNU/Linux http://www.debian.org/
GPG Public Key: 0x25BFB848. Please sign and encrypt your mail.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.2 (GNU/Linux)
Comment: Processed by Mailcrypt 3.5.8+ <http://mailcrypt.sourceforge.net/>
iD8DBQFDSrQrVcFcaSW/uEgRArtdAJ9BsCcU/zKzXXR/laDwP+zzw7pFpgCeIkuL
GbYTfPZLL9dyRmULQ49Y9YM=
=f//9
-----END PGP SIGNATURE-----
Tony Houghton
Roger Leigh <${rleigh}@invalid.whinlatter.ukfsn.org.invalid> wrote:
> I liked Python, for the small amount of time I spent with it. I just
> haven't fully grokked the function pass by sort-of-reference
> semantics.
>
> Unfortunately I need to be setuid, so I need a compiled language.
Couldn't you use sudo? Or you could probably even embed your python in a
minimal C wrapper.
Roger Leigh
Hash: SHA1
Tony Houghton <this.addre...@realh.co.uk> writes:
> In <874q7pj...@hardknott.home.whinlatter.ukfsn.org>,
> Roger Leigh <${rleigh}@invalid.whinlatter.ukfsn.org.invalid> wrote:
>
>> I liked Python, for the small amount of time I spent with it. I just
>> haven't fully grokked the function pass by sort-of-reference
>> semantics.
>>
>> Unfortunately I need to be setuid, so I need a compiled language.
>
> Couldn't you use sudo? Or you could probably even embed your python in a
> minimal C wrapper.
Not when the program in question (schroot) /is/ a sudo replacement,
with its own separate PAM configuration and access controls :)
- --
Roger Leigh
Printing on GNU/Linux? http://gimp-print.sourceforge.net/
Debian GNU/Linux http://www.debian.org/
GPG Public Key: 0x25BFB848. Please sign and encrypt your mail.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.2 (GNU/Linux)
Comment: Processed by Mailcrypt 3.5.8+ <http://mailcrypt.sourceforge.net/>
iD8DBQFDStIlVcFcaSW/uEgRAursAKDbH4WXpBaRaXaFKvi0QInlRIWM9QCg2CJB
VhgThucgJ01S4X1dW5tO07U=
=ceBR
-----END PGP SIGNATURE-----
Nix
> Python's really nice if you don't mind using an interpreted language.
> It's high level, so I find it a lot quicker and less fiddly than C (or
> C++, but I generally avoid that now), but it has the feel of a "real
> programming language", whereas perl feels more like a scripting
> language.
Downside: it's slow, slow, *slow*, even compared to Perl. A paltry
iteration through a ten thousand element list might take a second or so
on a P500.
Tony Houghton
Nix <nix-ra...@esperi.org.uk> wrote:
> On 09 Oct 2005, Tony Houghton prattled cheerily:
>> Python's really nice if you don't mind using an interpreted language.
>> It's high level, so I find it a lot quicker and less fiddly than C (or
>> C++, but I generally avoid that now), but it has the feel of a "real
>> programming language", whereas perl feels more like a scripting
>> language.
>
> Downside: it's slow, slow, *slow*, even compared to Perl. A paltry
> iteration through a ten thousand element list might take a second or so
> on a P500.
I think there are some (new?) list-like types optimised for certain
types of access.
Nix
> In <87achg8...@amaterasu.srvr.nix>,
> Nix <nix-ra...@esperi.org.uk> wrote:
>
>> On 09 Oct 2005, Tony Houghton prattled cheerily:
>>> Python's really nice if you don't mind using an interpreted language.
>>> It's high level, so I find it a lot quicker and less fiddly than C (or
>>> C++, but I generally avoid that now), but it has the feel of a "real
>>> programming language", whereas perl feels more like a scripting
>>> language.
>>
>> Downside: it's slow, slow, *slow*, even compared to Perl. A paltry
>> iteration through a ten thousand element list might take a second or so
>> on a P500.
>
> I think there are some (new?) list-like types optimised for certain
> types of access.
Any idea what they're called? I seem to be too stupid to find them
in the language reference...
Geoffrey Clements
news:87k6gk3...@amaterasu.srvr.nix...
> On 11 Oct 2005, Tony Houghton wrote:
>> In <87achg8...@amaterasu.srvr.nix>,
>> Nix <nix-ra...@esperi.org.uk> wrote:
>>
>>> On 09 Oct 2005, Tony Houghton prattled cheerily:
>>>> Python's really nice if you don't mind using an interpreted language.
>>>> It's high level, so I find it a lot quicker and less fiddly than C (or
>>>> C++, but I generally avoid that now), but it has the feel of a "real
>>>> programming language", whereas perl feels more like a scripting
>>>> language.
>>>
>>> Downside: it's slow, slow, *slow*, even compared to Perl. A paltry
>>> iteration through a ten thousand element list might take a second or so
>>> on a P500.
>>
>> I think there are some (new?) list-like types optimised for certain
>> types of access.
>
> Any idea what they're called? I seem to be too stupid to find them
> in the language reference...
>
mmmm - i don't think there are new list types but there are a number of
optimizations around list types and other places, see:
http://www.python.org/doc/2.4/whatsnew/node12.html#SECTION0001210000000000000000
--
Geoff
Tony Houghton
Nix <nix-ra...@esperi.org.uk> wrote:
> On 11 Oct 2005, Tony Houghton wrote:
>> In <87achg8...@amaterasu.srvr.nix>,
>> Nix <nix-ra...@esperi.org.uk> wrote:
[Python]
>>> Downside: it's slow, slow, *slow*, even compared to Perl. A paltry
>>> iteration through a ten thousand element list might take a second or so
>>> on a P500.
>>
>> I think there are some (new?) list-like types optimised for certain
>> types of access.
>
> Any idea what they're called? I seem to be too stupid to find them
> in the language reference...
Hm, I might have been overstating the case a bit now I look again. It
was in the tutorial I found them, section 11.7 (for python 2.4). There's
an array class, an itertools module and a collections module etc.
Mark Hobley
>
> Some examples:
>
> 1) Often when I yum update something breaks - often something difficult to
> fix
>
Stick with Debian Stable on a 2.6.x Kernel.
Don't update unless you need to.
> 8) SSH setup isn't consistent across linux distros - some do X11
> forwarding others dont - some set the DISPLAY variable others dont ...
> very fucking annoying if you have lots of machines to admin.
>
Standardise across the network. Use Debian Stable throughout.
> 9) The only thing constant about administering apache across distros is
> the name - every other fucking thing has a different path/filename...
>
Again, stick with Debian Stable.
> 10) Evolution is the best visual client - but you still cant drag and drop
> attachments from within KDE .... desktop integration pha !
>
I thought you said that you were using Gnome earlier.
> Im I alone in having these day to day problems
As a Windows user would you feel any better knowing that everyone else has
the same problems with the computer crashing ?
> things are getting worse not better as the months tick by.
Stick with what you like. Get rid of packages that you don't like.
If a package gets worse, use the old version, or create a fork.
Keep things simple. Use icewm instead of gnome.
Set up a network server, so that all your machines use the same binaries.
Regards,
Mark.
--
Mark Hobley
393 Quinton Road West
QUINTON
Birmingham
B32 1QE
Telephone: (0121) 247 1596
International: 0044 121 247 1596
Email: markhobley at hotpop dot donottypethisbit com
Nix
>> in the language reference...
>
> mmmm - i don't think there are new list types but there are a number of
> optimizations around list types and other places, see:
> http://www.python.org/doc/2.4/whatsnew/node12.html#SECTION0001210000000000000000
Yeah, well, even with those its speed leaves something to be desired.
(I was doing delta debugging with diffs broken down into individual
changed lines and applied in differing ways. It chewed for a *long* time
--- about three minutes --- constructing the initial list from a 20,000
line diff... after that the actual list manipulation was blindingly
fast, though.)