[Adium-devl] Let's switch to a DVCS

[Felipe Contreras]

On Thu, Apr 16, 2009 at 12:08 AM, Augie Fackler <lists at durin42.com> wrote:
>
> On Apr 15, 2009, at 4:04 PM, Felipe Contreras wrote:
>
>> On Wed, Apr 15, 2009 at 11:53 PM, Zachary West <zacwest at gmail.com> wrote:
>>>
>>> On Wed, Apr 15, 2009 at 16:45, Felipe Contreras
>>> <felipe.contreras at gmail.com> wrote:
>>>>
>>>> With git you can have the repository divided into different parts and
>>>> then set "graft points" to join them.
>>>>
>>>> Which means you can move to git right now, and forget about the old
>>>> history. The conversion of the old history can happen separately, and
>>>> at any given moment the people that want to have the whole history can
>>>> manually set "graft-points" to have it.
>>>>
>>>> If you make mistakes on the conversion (which you probably will) it
>>>> doesn't matter, because you would only be re-creating the old history
>>>> repo without affecting the new one, and you would simply need to
>>>> update the graft point.
>>>
>>> This sounds pretty cool, and definitely piques my interest. I don't
>>> really know enough to compare git to our chosen-by-the-masses (people
>>> other than me, heh) decision to use mercurial. If it's got the edge in
>>> size, and has the grafting which would let us have a relatively small
>>> checkout with recent history, but the ability to get all of the
>>> history…should we reconsider our plans for mercurial?
>>
>> It's strikes me as surprising that mercurial seems to be the chosen
>> one considering git is the clear winner in the DVCS battle. I would
>> gladly provide references, but for starters all of the big projects
>> that recently announced a move to a DVCS chose git.
>
> I'll take that bait: Python is moving to hg. All of the OpenSolaris stuff is
> in hg. Mozilla has embraced hg.
>
> I'll grant OpenSolaris as not huge, but you can't discount Python and
> Mozilla.

Yeah Python is moving to hg, is that really surprising? Python
choosing a DVCS that is implemented in Python. I don't think git was
given a fair chance there.

As for OpenSolaris I couldn't care less. I don't know anything of that
community.

And yeah the fact that Mozilla chose hg is important, that's why I
said *recently*. Mozilla switched a long time, before git was the
clear winner.

For a quick list of big projects that recently chose git: GNOME, Perl,
Qt, GStreamer, Pulseaudio, OpenEmbedded, Farsight, and KDE is probably
on the way.

Also, ALSA moved from hg to git, perhaps Mozilla will do too at some point?

Also, here's a tale of a guy that realized some problems with hg while
doing the KDE conversion to hg... in the process he decided to switch
to git:
http://labs.trolltech.com/blogs/2007/12/28/git-import-of-kde-or-a-tale-of-how-i-am-lazy/

>> I am clearly biased but I could explain why I think mercurial has
>> serious design issues that will never be solved, and git of course
>> doesn't. But only if somebody is interested :)
>
> Git has its own design issues. Frankly, this isn't the forum for discussing
> that. Git and hg made different tradeoffs with slightly different goals in
> mind, and over time, they're approaching each other asymptotically.

You could say in general hg performance is getting closer to git's
performance and will never reach it, but that's not so relevant. The
two different design makes certain operations faster on hg, while
others are faster on git, but IMHO the operations that are most
commonly used are faster on git, and the ones that are faster on hg
are essentially irrelevant. That will not change, ever.

Cheers.

-- 
Felipe Contreras