On Tue, 2003-09-02 at 23:17, Jon Smirl wrote:
--- Michel Dnzer <[EMAIL PROTECTED]> wrote:
On Tue, 2003-09-02 at 19:37, Jon Smirl wrote:
The BitKeeper people have said they would love to be the host for DRI/Mesa source.
I strongly object to this, I won't take part in development with non-free tools. If CVS really is a problem, there are free alternatives which should be more than powerful enough for the DRI tree. The merge arguments are moot, see http://www.mail-archive.com/[EMAIL PROTECTED]/msg02472.html .
BK is free to download. The free license has the quirky condition that you
can't also develop a competing source code control tool.
This and other restrictions make it very non-free as in freedom, unacceptably so for me.
There are probably a thousand people using BK to work on the kernel right now. A few people working on the kernel had a conflict with this condition. Some bought a BK license which removes the restriction, others are using a read-only CVS gateway and mailing their patches to a person using BK.
As Linus said, this works for the Linux kernel because people always had to send him patches anyway. We have different requirements. If nothing else, bk is overkill for the DRI IMO.
A side effect of using BK is that they are providing free hosting for BK projects on a high bandwidth server. SourceForge is really the main problem.
Exactly, so why are you trying to sneak in a non-free tool instead of addressing the real problem?
Wow, instant flame war.
Anyway, let's take a step back - any change on the magnitude of shifting to a different version control system would have to be by consensus of active developers and it's clear now that BK won't get that, so let's just put that idea away and move on.
PS - If you tried using BK for a while you'd quickly see why it is so much better than CVS. Peer to peer is a much better solution that CVS' client/server model. The BK architecture is 10 years ahead of CVS'. Several projects are trying to clone BK but none are close to being complete.
I'm pretty sure Subversion does at least what we need, and Xouvert is even using arch, though admittedly it remains to be seen how that works out.
What is wrong right now is that we can't get uptodate versions of the repository to non-developers.
Non-developers currently don't have BK, arch, subversion or even cvsup installed. They do seem to often have CVS installed, and we've successfully used it in the past to iterate test-debug-patch cycles. I don't think we can expect them to be installing a new version tool to participate in that existing process.
A lot of them do have rsync, but even that wasn't installed by default in redhat until pretty recently. It's also a lot less rich than cvs.
So, anonymous cvs is probably our best bet for rapid distribution of changes etc to users.
This means two things
1) we need to fix anonymous CVS now
2) the discussion of developer version control system doesn't impact the fact that we need to fix anonymous CVS access for non-developers, so is not to be considered in any way urgent, imminent or even likely.
If we ever did move to a different system, I think we'd want to keep an anonymous CVS access for non-developer access.
So, right now, the urgent thing is to find a reliable site to host our cvs, including pserver anonymous cvs.
No need for a flame war, talk of other version systems is basically idle banter and healthy venting of frustration. I expect we'll be using cvs for some time to come, if not longer.
Keith
------------------------------------------------------- This sf.net email is sponsored by:ThinkGeek Welcome to geek heaven. http://thinkgeek.com/sf _______________________________________________ Dri-devel mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/dri-devel
