Hi Marcin, On Sep 03, 2014, at 08:51 AM, Marcin Kulisz wrote: >Barry I see you're in favour of dgit but there is one problem with it (at >least was last time I checked) it's not going to work for people who aren't >DDs and this should be kept in mind if you want to attract none DDs to join >DPMT or/and PAPT. I wouldn't say I'm in favor of dgit, but I'm certainly curious! :) I've only briefly played with it and not for anything 'real'. There's not a lot of information available about dgit afaict, and I'm still waiting for the DC14 video on that session to be posted. I will say that I am philosophically aligned with the aims of dgit, due to my fondness of and familiarity with Ubuntu Distributed Development (UDD), which IIUC, is essentially the same idea s/Debian/Ubuntu/ and s/git/bzr/. UDD though has lots of technical and operational problems, limited development resources, and narrow mindshare. IOW, a great idea that never took off. What I like about this approach is that you know the vcs always matches the archive[+]. In our current svn-based workflow (and adoption of gbp or git-dpm won't change this), it's possible for people to do `apt-get source`, hack, and upload and introduce a discrepancy between the archive and the vcs. I don't know of any technical way to prevent this, but thankfully it doesn't happen too often, especially with team packages. I've encountered this once or twice and it's a pain to reconstruct the vcs. Maybe it would be easier with git, where there seem to be better tools to back-fill history[*]. Your point about the limitations of dgit to DDs is an important one. I personally think teamwork is a fantastic way for new contributors to build a good reputation on the path to DM or DD, and I wouldn't want to adopt any practice that locks out new contributors. On top of that, because of git's distributed nature, I think it will provide an even better way of getting contributions from non-DDs, and I definitely want to preserve that. Cheers, -Barry [+] That's in fact one of the biggest problems with UDD. Due to the high rate of importer failures, the vcs very often does *not* match the archive, defeating the whole purpose of the technique! I'm told by the smart folks that git would be much less problematic than bzr here. [*] As an example, I just uploaded a new python-colorama, which has been essentially abandoned by its MIA maintainer. With permission of the MIA team, I've adopted that into DPMT, although I haven't moved its git repo yet, and won't until we know what we're doing with git. But it was missing some pristine tar and upstream history, and it wasn't too hard to back-fill this. Now whether I did that correctly or not... ;)
Attachment:
signature.asc
Description: PGP signature