Mono
Fergus Henderson
good or bad for free software?
--
Fergus Henderson <f...@cs.mu.oz.au> | "I have always known that the pursuit
The University of Melbourne | of excellence is a lethal habit"
WWW: <http://www.cs.mu.oz.au/~fjh> | -- the last words of T. S. Garp.
brl...@my-deja.com
> Is the Mono project <http://ximian.com/tech/mono-index.php3>
> good or bad for free software?
Bad. We should not help the adoption of a MSFT-designed user
authentication system, which they will surely embrace-and-extend as they
did with Kerberos. Effort would be better spent on a high-usability
public-key infrastructure (PKI).
So far, the CLR is no better than the JVM for real multi-language
support, e.g. Scheme-style tail-call optimization and continuations.
Tail-call optimization is at least "planned for a future version", but
I'm from Missouri on MSFT promises: Show me.
As has been the case with Java class libraries, free OSes will always be
playing catch-up for C# class libraries. Only this time it won't be
catch-up with Solaris and Windows. It will just be Windows.
--
Bruce R. Lewis http://brl.sourceforge.net/
I rarely read mail sent to brl...@my-deja.com
John Hasler
> Is the Mono project <http://ximian.com/tech/mono-index.php3> good or bad
> for free software?
Considering that we can't even produce a stable Web browser, yes, starting
such a huge, complex project is very bad.
--
John Hasler
jo...@dhh.gt.org
Dancing Horse Hill
Elmwood, Wisconsin
Jeffrey Siegal
> Considering that we can't even produce a stable Web browser, yes, starting
> such a huge, complex project is very bad.
I don't find anything unstable about Mozilla 0.9.2. I've run both the
Windows and Linux versions for days at a time without problems.
Phillip Lord
Jeffrey> John Hasler wrote:
>> Considering that we can't even produce a stable Web browser, yes,
>> starting such a huge, complex project is very bad.
Jeffrey> I don't find anything unstable about Mozilla 0.9.2. I've
Jeffrey> run both the Windows and Linux versions for days at a time
Jeffrey> without problems.
Our experiences differ then. All the versions of
netscape or mozilla that I have ever tried have been up and down
quicker than Clinton's flies.
There are other huge projects that are quite stable
so I don't think that this reflects on anything, including Mono.
Phil
Jeffrey Siegal
> Our experiences differ then. All the versions of
> netscape or mozilla that I have ever tried have been up and down
> quicker than Clinton's flies.
Netscape is a very different story, at least on Linux. (The Windows
version is quite stable; I run it for weeks on end.) In its standard
configuration it is just a disaster. If you turn off Java and
JavaScript, you can do a bit better, but then you don't have much of a
browser.
Have you tried Mozilla 0.9.1 or 0.9.2? I find the stability to be
*vastly* improved over earlier versions.
John Hasler
> I don't find anything unstable about Mozilla 0.9.2.
I try Mozilla every month or so. Right now I have 0.9.1 installed. It
crashed several times in a couple of hours of use, can't display the Gimp
manual, gives me no control over the display of images, and has several
other bugs that I disremember just now. Galeon and Skipstone are slightly
different, but no better. Konqueror is more stable than Mozilla, but still
crashes about as often as Navigator and also has no useful image control.
I'll try Mozilla again in a few months, but my hopes aren't very high.
--
John Hasler
jo...@dhh.gt.org (John Hasler)
Dancing Horse Hill
Elmwood, WI
John Hasler
> Netscape is a very different story, at least on Linux. (The Windows
> version is quite stable; I run it for weeks on end.) In its standard
> configuration it is just a disaster.
I run 4.75 on Debian. It crashes several times a day, but at least it lets
me control images (I'm on a 28.8 dialup).
> If you turn off Java and JavaScript, you can do a bit better, but then
> you don't have much of a browser.
I turn JavaScript on about once a month (and Java never). If you have a
Web page that requires it you don't have much of a Web page.
John Hasler
> There are other huge projects that are quite stable...
What of a similar scale and complexity?
> ...so I don't think that this reflects on anything, including Mono.
Well, it appears to be being proposed by some of the people behind Gnome
which, while certainly a useful and admirable effort, is not exactly a
shining example of stability.
Jeffrey Siegal
> I try Mozilla every month or so. Right now I have 0.9.1 installed.
0.9.2 is much better. It was primarily crash-fixing update, with over
20 different crashes fixed.
It does still seem to have some problems with controlling images in
general, but its site-specific image blocking works well.
Jonathan Thornburg
John Hasler <jo...@dhh.gt.org> wrote:
>I try Mozilla every month or so. Right now I have 0.9.1 installed. It
>crashed several times in a couple of hours of use, can't display the Gimp
>manual, gives me no control over the display of images, and has several
>other bugs that I disremember just now. Galeon and Skipstone are slightly
>different, but no better. Konqueror is more stable than Mozilla, but still
>crashes about as often as Navigator and also has no useful image control.
Without defending any of these programs for their crashes, I should
note that the quasi-commercial Netscape Navigator crashes regularly
on every system where I've used it (with only occasional web browsing,
I'm probably averaging a crash every 2-3 days lately), including
various versions of Red Hat GNU/Linux, OSF/1, SunOS, OpenBSD, and
Irix. And certain web pages will reliably/reproducibly crash it.
On the positive side, I've *never* had lynx crash. (It is, admittedly,
a much smaller and simpler piece of software.) I just wish it could
render HTML tables into ASCII in a reasonable manner...
--
-- Jonathan Thornburg <jth...@thp.univie.ac.at>
Max-Planck-Institut fuer Gravitationsphysik (Albert-Einstein-Institut),
Golm, Germany http://www.thp.univie.ac.at/~jthorn/home.html
"The first strike in the American Colonies was in 1776 in Philadelphia,
when [...] carpenters demanded a 72-hour week." -- Anatole Beck
Phillip Lord
John> Phil writes:
>> There are other huge projects that are quite stable...
John> What of a similar scale and complexity?
GCC (with all the various front ends). Emacs (esp. if you
include the numerous add on packages).
>> ...so I don't think that this reflects on anything, including
>> Mono.
John> Well, it appears to be being proposed by some of the people
John> behind Gnome which, while certainly a useful and admirable
John> effort, is not exactly a shining example of stability. --
I've found that Gnome is getting better rapidly. The early
versions were somewhat flaky. Recently I've moved from using a RedHat
7.1 system to a RedHat 6 system (I am aware that this is
backwards!). The change in Gnome stability was quite noticeable.
Phil
Jeffrey Siegal
Jeffrey Siegal
> GCC (with all the various front ends). Emacs (esp. if you
> include the numerous add on packages).
There are others that can be added to the list. X11 and of course Linux
come to mind right away.
Kalle Olavi Niemitalo
> On the positive side, I've *never* had lynx crash. (It is, admittedly,
> a much smaller and simpler piece of software.) I just wish it could
> render HTML tables into ASCII in a reasonable manner...
Have you tried "links"? It has lynx-like key bindings but also
drop-down menus, can open a new "screen" or "xterm" and share history
lists, downloads in the background, and supports SSL, tables and
frames. Unfortunately, the version I have (0.95, IIRC) cannot
authenticate to the server and forcefully disables XON/XOFF with the
terminal. Some older versions used to crash too, but those problems
seem to have been fixed.
phil hunt
>f...@cs.mu.oz.au (Fergus Henderson) writes:
>
>> Is the Mono project <http://ximian.com/tech/mono-index.php3>
>> good or bad for free software?
>
>Bad. We should not help the adoption of a MSFT-designed user
>authentication system, which they will surely embrace-and-extend as they
>did with Kerberos. Effort would be better spent on a high-usability
>public-key infrastructure (PKI).
Why? What would this be useful for?
>So far, the CLR is no better than the JVM for real multi-language
>support, e.g. Scheme-style tail-call optimization and continuations.
>Tail-call optimization is at least "planned for a future version", but
>I'm from Missouri on MSFT promises: Show me.
Presumably the open source version could get ahead of MS on this,
then?
>As has been the case with Java class libraries, free OSes will always be
>playing catch-up for C# class libraries. Only this time it won't be
>catch-up with Solaris and Windows. It will just be Windows.
One problem with the JVM is you are in the Java environment, not
the Posix environment or whatever. What is the CLR like here?
Does it have its own environment? its own windowing library?
--
## Philip Hunt ## ph...@comuno.freeserve.co.uk ##
phil hunt
>Fergus Henderson writes:
>> Is the Mono project <http://ximian.com/tech/mono-index.php3> good or bad
>> for free software?
>
>Considering that we can't even produce a stable Web browser, yes, starting
>such a huge, complex project is very bad.
konqueror seems to be doing quite well, IMO
brl...@my-deja.com
> On 10 Jul 2001 09:01:27 -0400, brl...@my-deja.com
> <brl...@my-deja.com> wrote:
> >Effort would be better spent on a high-usability
> >public-key infrastructure (PKI).
>
> Why? What would this be useful for?
It would be useful for identifying yourself to web sites without
divulging secret info or depending on one centralized service like
Passport. Look at SSL certificates: These let you choose an authority
to get a certificate from, but once you have that certificate you can
interact with other servers without further intervention by the
certificate authority, and you never give info to the other servers that
would let them impersonate you to another server.
Usability is the reason SSL certificates never took off. The dialog
boxes involved in getting your initial cert were too confusing. Setting
up SSL certificate authentication on an apache server involves too much
configuration. So people just go with separate password-protected
accounts on a plethora of web sites.
A moderate effort to increase the usability of SSL certificates would
take the wind out of Passport's sails.
> >So far, the CLR is no better than the JVM for real multi-language
> >support, e.g. Scheme-style tail-call optimization and continuations.
> >Tail-call optimization is at least "planned for a future version", but
> >I'm from Missouri on MSFT promises: Show me.
>
> Presumably the open source version could get ahead of MS on this,
> then?
People might be reluctant to do an implementation of, say,
continuations, before a standard specifies what they should look like.
> >As has been the case with Java class libraries, free OSes will always be
> >playing catch-up for C# class libraries. Only this time it won't be
> >catch-up with Solaris and Windows. It will just be Windows.
>
> One problem with the JVM is you are in the Java environment, not
> the Posix environment or whatever. What is the CLR like here?
> Does it have its own environment? its own windowing library?
For WORA to work, the runtime has little choice but to define its own
environment and windowing library.
John Hasler
> Without defending any of these programs for their crashes, I should note
> that the quasi-commercial Netscape Navigator crashes regularly on every
I use Navigator 4.75 under Debian regularly. It crashes several times a
day. I don't see that the poor quality of most consumer grade commercial
software is relevant, though.
> on the positive side, I've *never* had lynx crash.
Unfortunately, the majority of Web pages are so poorly designed as to be
unuseable with text browsers.
John Hasler
> What of a similar scale and complexity?
Phil writes:
> GCC (with all the various front ends). Emacs (esp. if you include the
> numerous add on packages).
I think that the proposed Mono project is at least an order of magnitude
more complex than either of those, and neither was as radically new as
Mono.
Also, consider how long gcc and emacs have taken to reach their present
state. Is Mono going to take as long?
Phillip Lord
John> I wrote:
>> What of a similar scale and complexity?
John> Phil writes:
>> GCC (with all the various front ends). Emacs (esp. if you include
>> the numerous add on packages).
John> I think that the proposed Mono project is at least an order of
John> magnitude more complex than either of those
I'm not sure that I agree with this. Emacs for instance
is now an enormous piece of software (if you include all the bits and
pieces). I have for instance 50M of lisp in my home space. This is not
including the main emacs install, which is a lot of source.
John> and neither was as radically new as Mono.
Maybe so....
John> Also, consider how long gcc and emacs have taken to reach
John> their present state. Is Mono going to take as long?
For mono to get as mature as something like emacs or gcc
yes probably. Obviously the notion of what is mature and what is not
is a matter of opinion.
Phil
John Hasler
> Considering that we can't even produce a stable Web browser, yes, starting
> such a huge, complex project is very bad.
Philip Hunt writes:
> konqueror seems to be doing quite well, IMO
It doesn't crash as often as Navigator and Mozilla, but it still crashes.
I'd use it, but it has no useful image control.
Fergus Henderson
>Fergus Henderson writes:
>> Is the Mono project <http://ximian.com/tech/mono-index.php3> good or bad
>> for free software?
>
>Considering that we can't even produce a stable Web browser, yes, starting
>such a huge, complex project is very bad.
Hmm, if stability is such a problem, then the solution might be to
develop a type-safe and memory-safe infrastructure, where you have
automatic memory management and the system does array bounds checking
and catches any attempts to access uninitialized variables, and with
support for exception handling. It would also be nice if it was easier
to write appropriate parts of the software in higher-level languages.
Oh, wait, I think someone is already working on that...
Fergus Henderson
>So far, the CLR is no better than the JVM for real multi-language
>support, e.g. Scheme-style tail-call optimization and continuations.
>Tail-call optimization is at least "planned for a future version", but
>I'm from Missouri on MSFT promises: Show me.
I think you may have misunderstood the message.
The EMCA specs includes support for marking tail calls using the "tail.call"
instruction, and this is implemented in Microsoft's implementation.
They work; you can use tail calls, and they won't consume stack space.
The problem is just that in Microsoft's current implementation,
tail calls are SLOW, quite a bit slower than ordinary calls,
rather than being faster than ordinary calls, as they ought to be.
That's why MS still need to work on optimizing tail calls,
even though the infrastructure already provides support for tail calls.
The CLR is definitely better than the JVM for real multi-language
support, IMHO. As well as tail calls, it has by-reference arguments,
value types (i.e. stack-allocated structs, as opposed to heap-allocated
classes), unsigned arithmetic, dynamic stack allocation. and support
for unverifiable code (including pointer arithmetic and function pointers).
These things tend to be difficult to emulate efficiently on the JVM.
Linus Torvalds
John Hasler <jo...@dhh.gt.org> wrote:
>
>Philip Hunt writes:
>> konqueror seems to be doing quite well, IMO
>
>It doesn't crash as often as Navigator and Mozilla, but it still crashes.
>I'd use it, but it has no useful image control.
I started using konqueror because it's the only thing that renders
things well (anti-aliased fonts etc are a major issue for me, I cringe
every time I see a regular netscape page these days).
It's been getting a lot better too, and what really impressed me about
the whole thing was how well the KDE people react to bug-reports. I had
a site that I reported as badly rendered, and three days later the
problem was 100% fixed.
You really have to be on the CVS tree to see konqueror shine, and it's a
bit painful to do that because of all the dependencies (you have to make
sure that you update all the relevant projects so that you don't have
old plugins etc messing up), but I'm completely sold on it. Not only
does it look better than Mozilla and Navigator, but it doens't force
that silly side-bar on you, and the developers seem to be a-ok.
(I also much prefer koffice over StarOffice - I hate the stupid "we know
what kind of desktop you like" approach SO has. I've not seen any of
the upcoming OpenOffice to see if they really got that right, but it's
supposed to be fixed).
Linus
phil hunt
>
>Usability is the reason SSL certificates never took off. The dialog
>boxes involved in getting your initial cert were too confusing. Setting
>up SSL certificate authentication on an apache server involves too much
>configuration. So people just go with separate password-protected
>accounts on a plethora of web sites.
>
>A moderate effort to increase the usability of SSL certificates would
>take the wind out of Passport's sails.
It's interesting you say that, ads my analysis is that encrypted email
hasn't taken off for essentially the same reason, that PGP and GnuPG
are to difficult to use. Hence I am currently writing a program which
will give people effort-free public key encryption.
>> >So far, the CLR is no better than the JVM for real multi-language
>> >support, e.g. Scheme-style tail-call optimization and continuations.
>> >Tail-call optimization is at least "planned for a future version", but
>> >I'm from Missouri on MSFT promises: Show me.
>>
>> Presumably the open source version could get ahead of MS on this,
>> then?
>
>People might be reluctant to do an implementation of, say,
>continuations, before a standard specifies what they should look like.
Possibly the open source version of CLR could implement them anyway,
(with a flag so that incompatible extensions can be turned off if desired)
regardless of whether MS has, and in that way we could "embrace
and extend" MS?
phil hunt
>I wrote:
>> What of a similar scale and complexity?
>
>Phil writes:
>> GCC (with all the various front ends). Emacs (esp. if you include the
>> numerous add on packages).
>
>I think that the proposed Mono project is at least an order of magnitude
>more complex than either of those, and neither was as radically new as
>Mono.
The KDE and GNOME projects are hardly simple, yet they've come a
long way in a short time.
Dylan Thurston
>> on the positive side, I've *never* had lynx crash.
>
>Unfortunately, the majority of Web pages are so poorly designed as to be
>unuseable with text browsers.
With w3m, my favorite text browser, I am able to see almost all pages
I'm interested in (which does not contradict John Hasler's comment, of
course...) It has good frames support, but no JavaScript. And it does
its best with inline images, but there are unavoidable problems.
--Dylan Thurston
d...@math.harvard.edu
John Hasler
> You really have to be on the CVS tree to see konqueror shine,...
I'm on a 28.8 dialup on a shared phone line.
brl...@my-deja.com
> I think you may have misunderstood the message.
> The EMCA specs includes support for marking tail calls using the "tail.call"
> instruction, and this is implemented in Microsoft's implementation.
> They work; you can use tail calls, and they won't consume stack space.
I stand corrected. I may be remembering a "future plan" type of message
from last year. Couldn't find it on google, though I found
comp.lang.scheme threads on the topic from January and April.
Fergus Henderson
>It's been getting a lot better too, and what really impressed me about
>the whole thing was how well the KDE people react to bug-reports. I had
>a site that I reported as badly rendered, and three days later the
>problem was 100% fixed.
Somehow that doesn't surprise me...
but it would be an interesting experiment for you to try posting your
next bug report as "jrando...@aol.com (Kool Hackr Dude)" rather than as
"torv...@penguin.transmeta.com (Linus Torvalds)" and see if you get the
same response time ;-)
phil hunt
>torv...@penguin.transmeta.com (Linus Torvalds) writes:
>
>>It's been getting a lot better too, and what really impressed me about
>>the whole thing was how well the KDE people react to bug-reports. I had
>>a site that I reported as badly rendered, and three days later the
>>problem was 100% fixed.
>
>Somehow that doesn't surprise me...
>but it would be an interesting experiment for you to try posting your
>next bug report as "jrando...@aol.com (Kool Hackr Dude)" rather than as
ITYM "Kewl h4x0r d00d" :-)
>"torv...@penguin.transmeta.com (Linus Torvalds)" and see if you get the
>same response time ;-)
--