On Tue, 21 Nov 2006 08:36:53 -0800, Jeffrey Cunningham jeffrey@cunningham.net wrote:
On Tue Nov 21, 2006 at 08:12:44AM +0100, Edi Weitz wrote:
Does usage of plain old CVS hold back SLIME development in any way?
Many big companies have their firewalls set up to block the port CVS uses - I've been told the MSEE training their IT people get tells them to do this. That means you can't grab CVS versions of Slime behind these firewalls.
If you just want the current CVS version to /use/ SLIME, you can get it through http, of course:
http://common-lisp.net/cgi-bin/viewcvs.cgi/root.tar.gz?root=slime&view=t...
But, yes, if you actually want to work on SLIME and submit code, this might be a problem. Does anyone know how many active developers we have who are paid by their big company employers to hack SLIME during their working hours?
:)
But wait, project members use CVS via ssh, don't they? So, if the CVS port is blocked, that shouldn't be problem.
I assume I'm not alone with this problem.
Darcs just works.
For some people... :)
On Tue Nov 21, 2006 at 05:47:21PM +0100, Edi Weitz wrote:
If you just want the current CVS version to /use/ SLIME, you can get it through http, of course:
Yes - that's how I do get it at work, and grateful I am that its available that way, too!
But wait, project members use CVS via ssh, don't they? So, if the CVS port is blocked, that shouldn't be problem.
Unfortunately - at least where I work - the same mentality that blocks the CVS port also blocks the SSH port. VPN is the only access allowed and that's very difficult to set up on Linux. My associates at work all use something called Aventail on Windows - and the moan and complain about how awful it all the time.
I'm just throwing these comments in for completeness. I happen to like Darcs, but I'll be happy to work with whatever you guys want to use (as long as its not Aventail ;-)
Regards, -Jeff
I would like to warn that darcs has some definite performance problems, allthough slime is so small that it is unlikely to be a problem.
After having sufficiently tired with CVS, I went on a quest to find a more moderne replacement. I arrived, at darcs after an uneasy period using TLA/Arch, following a number of excited reviews in the lisp community.
However, I have twice (with different projects) experienced that merging between repositories stopped working. Darcs uses some hairy lazy-evaluation analysis of the complete patch combination graph to determine dependencies and this is what broke, I think. In one instance it just kept on working, and after 23 hours I lost patience and gave up, whether it was due to an endless loop in the code or a combinatorial explosion in the data I shall not say, but work it did not.
In fairness, one example was my homediretory where I keep dot files under revision control to exercise control and to enable safe moving of information between my various accounts. It is rather impressive how many files and directories that accumulates in ones homedir over time, so this was clearly putting some stress on darcs. The other exmple involved a number of binary files which may have contributed to the problem.
As I said, I do not think that slime is in any direct peril but those experiences made it hard for me to trust darcs, all its other nice features aside.
I am now a happy subversion user. One of the strong points for subversion in my book is that it is developed by some of the old CVS notabilities.
------------------------+----------------------------------------------------- Christian Lynbech | christian #@ defun #. dk ------------------------+----------------------------------------------------- Hit the philistines three times over the head with the Elisp reference manual. - petonic@hal.com (Michael A. Petonic)
As I said, I do not think that slime is in any direct peril but those experiences made it hard for me to trust darcs, all its other nice features aside.
this is due to an exponential algorithm in darcs, but it only happens when darcs tries to merge some badly nested conflicts. it only happens in "normal" usage when you have branches that get too far from each other by touching the same files at the same positions too much.
this is a structural weakness of darcs, but the good features come from this internal operation, too.
the guys at #darcs said last time i ran into this problem that they are planning to bail out from such an exponential situation and mark the conflict for user resolution (which would happen anyway after 10^n years... :)
haven't seen this change yet, though.
One of my main reservations about darcs is that slime is such an important part of a modern lisp environment, and it has a tendency to avoid releases, rather relying on the latest version from the repository to serve as the "version to be used", that putting the canonical version in darcs is likely to cause a decent number of users to stay away from it because 1) they don't have darcs or 2) they don't know or understand darcs. I would imagine there are still a reasonable number of platforms for which darcs doesn't exist yet. Telling first time lisp users (who, IMHO, should be using slime) to go find a haskell compiler and to build darcs seems a bit rough. Perhaps lisp-boxes could be built with darcs, etc... for various platforms, but I think sticking to a more "mainstream" repository like CVS or SVN is going to help with the ongoing adoption of slime.
Of course I have no data to back up my claims and this is all speculation, but I think it's worth considering. I would also claim that while darcs has some nice bells and whistles, the status quo works reasonably well and most folks have figured out how to deal with the trees to create small, manageable patches for inclusion into the try. Educating the developer base to do so without requiring darcs is probably going to require an overall lesser amount of work than trying to convince new users to use darcs (or moving to a more release-oriented distribution model).
Cyrus
On Nov 23, 2006, at 1:53 AM, Christian Lynbech wrote:
I would like to warn that darcs has some definite performance problems, allthough slime is so small that it is unlikely to be a problem.
After having sufficiently tired with CVS, I went on a quest to find a more moderne replacement. I arrived, at darcs after an uneasy period using TLA/Arch, following a number of excited reviews in the lisp community.
However, I have twice (with different projects) experienced that merging between repositories stopped working. Darcs uses some hairy lazy-evaluation analysis of the complete patch combination graph to determine dependencies and this is what broke, I think. In one instance it just kept on working, and after 23 hours I lost patience and gave up, whether it was due to an endless loop in the code or a combinatorial explosion in the data I shall not say, but work it did not.
In fairness, one example was my homediretory where I keep dot files under revision control to exercise control and to enable safe moving of information between my various accounts. It is rather impressive how many files and directories that accumulates in ones homedir over time, so this was clearly putting some stress on darcs. The other exmple involved a number of binary files which may have contributed to the problem.
As I said, I do not think that slime is in any direct peril but those experiences made it hard for me to trust darcs, all its other nice features aside.
I am now a happy subversion user. One of the strong points for subversion in my book is that it is developed by some of the old CVS notabilities.
+----------------------------------------------------- Christian Lynbech | christian #@ defun #. dk
+----------------------------------------------------- Hit the philistines three times over the head with the Elisp reference manual. - petonic@hal.com (Michael A. Petonic) _______________________________________________ slime-devel site list slime-devel@common-lisp.net http://common-lisp.net/mailman/listinfo/slime-devel
On 23-Nov-06, at 2:12 PM, Cyrus Harmon wrote:
I would imagine there are still a reasonable number of platforms for which darcs doesn't exist yet.
As someone who follows along in the development of Slime -- and someone who gets regular updates via CVS -- I can say that I wouldn't be very enthusiastic about switching to darcs. I spend a lot of time in Slime on a Macbook Pro and neither Fink nor MacPorts (formerly DarwinPorts) supports an installation of the Haskell compiler on the Intel architecture yet.
If I had to I would go and build the Haskell compiler myself, but this is not a course of action I would see as favourable. It most certainly would not be favourable to other people I work with who don't spend as much time using Lisp who needed help setting up Slime in the first place.
I would imagine there are still a reasonable number of platforms for which darcs doesn't exist yet.
As someone who follows along in the development of Slime -- and someone who gets regular updates via CVS -- I can say that I wouldn't
fyi, seems like i was the only one pushing it, but i already gave up. i'll keep my branch in a publically available darcs repo, though.
darcs binaries: http://darcs.net/DarcsWiki/CategoryBinaries
On 23/11/06, Geoff Wozniak geoff@wozniak.ca wrote:
On 23-Nov-06, at 2:12 PM, Cyrus Harmon wrote:
I would imagine there are still a reasonable number of platforms for which darcs doesn't exist yet.
As someone who follows along in the development of Slime -- and someone who gets regular updates via CVS -- I can say that I wouldn't be very enthusiastic about switching to darcs. I spend a lot of time in Slime on a Macbook Pro and neither Fink nor MacPorts (formerly DarwinPorts) supports an installation of the Haskell compiler on the Intel architecture yet.
The PPC darcs binary runs well under Rosetta, I use it all the time on my Macbook pro.
Cheers Brad
2006/11/24, Brad Beveridge brad.beveridge@gmail.com:
On 23/11/06, Geoff Wozniak geoff@wozniak.ca wrote:
On 23-Nov-06, at 2:12 PM, Cyrus Harmon wrote:
I would imagine there are still a reasonable number of platforms for which darcs doesn't exist yet.
As someone who follows along in the development of Slime -- and someone who gets regular updates via CVS -- I can say that I wouldn't be very enthusiastic about switching to darcs. I spend a lot of time in Slime on a Macbook Pro and neither Fink nor MacPorts (formerly DarwinPorts) supports an installation of the Haskell compiler on the Intel architecture yet.
The PPC darcs binary runs well under Rosetta, I use it all the time on my Macbook pro.
Why? There is a binary of darcs distributed for OSX/Intel. It's in the Darcs Wiki. It works perfectly for me.
The other alternative is installing via fink, but there you have to switch to the unstable branch and build the Haskell compiler, and this takes ages.
Juanjo
On 24/11/06, Juan Jose Garcia-Ripoll jjgarcia@users.sourceforge.net wrote:
2006/11/24, Brad Beveridge brad.beveridge@gmail.com:
On 23/11/06, Geoff Wozniak geoff@wozniak.ca wrote:
On 23-Nov-06, at 2:12 PM, Cyrus Harmon wrote:
I would imagine there are still a reasonable number of platforms for which darcs doesn't exist yet.
As someone who follows along in the development of Slime -- and someone who gets regular updates via CVS -- I can say that I wouldn't be very enthusiastic about switching to darcs. I spend a lot of time in Slime on a Macbook Pro and neither Fink nor MacPorts (formerly DarwinPorts) supports an installation of the Haskell compiler on the Intel architecture yet.
The PPC darcs binary runs well under Rosetta, I use it all the time on my Macbook pro.
Why? There is a binary of darcs distributed for OSX/Intel. It's in the Darcs Wiki. It works perfectly for me.
Thanks, it had been a while since I updated - the Intel build must be new.
Cheers Brad
Jeffrey Cunningham jeffrey@cunningham.net writes:
Unfortunately - at least where I work - the same mentality that blocks the CVS port also blocks the SSH port. VPN is the only access allowed and that's very difficult to set up on Linux.
Usually you can tunnel ssh out via e.g. a HTTPS proxy and then tunnel everything else via ssh. Very few corporate firewalls can contain a suitably motivated person! Google knows many tools.
Luke Gorrie luke@synap.se writes:
Usually you can tunnel ssh out via e.g. a HTTPS proxy and then tunnel everything else via ssh. Very few corporate firewalls can contain a suitably motivated person! Google knows many tools.
Yes, this is usually possible. But you lose plausible deniability. If you're tunnelling ssh over udp dns requests because a corporate firewall blocks all other attempts at getting out, you can hardly claim ignorance when They confront you with breach of coporate rules and/or contract. That may or may not be a firing offence, depending on location and employer.
If you have as part of your contract "Thou shalt not ssh to external machines", in theory don't even need to block port 22 out.
So, while technically possible, it may not be a good idea.
...Peder...
-
Attila Lendvai -
Brad Beveridge -
Christian Lynbech -
Cyrus Harmon -
Edi Weitz -
Geoff Wozniak -
Jeffrey Cunningham -
Juan Jose Garcia-Ripoll -
Luke Gorrie -
peder＠klingenberg.no