[Adium-devl] Version control

[Chris Forsythe]

On Oct 8, 2007, at 3:50 AM, Colin Barrett wrote:

>
> On Oct 8, 2007, at 1:22 AM, Andreas Monitzer wrote:
>
>> On Oct 08, 2007, at 06:25, Colin Barrett wrote:
>>
>>> 1) We should move Adium to /some/ sort of distributed version  
>>> control
>>> system.
>>>
>>> DVCS is the way of the future.
>>
>> That shouldn't be an argument at all. Just because it's hip right at
>> the moment doesn't mean that you should jump ship to the buzzword of
>> the day.
>
> We've done this before, in 2005, before "distributed" was even a
> buzzword. A number of us tried out svk, but it didn't really work all
> that well (we couldn't get it actually publish our changes, iirc)
>


If it's the way of the future, it'll be there when we get to it. I'm  
of the opinion that we should at least wait until svn 1.5, and see if  
that solves our needs.

Which comes to my really basic question. What are our needs? Make a  
list, then we review each versioning system. If svn doesn't do what  
we need, that's fine, but let's pick one that does what we need. Keep  
in mind that we have to retrain hundreds of users who pull svn  
nightly as well. The point that Andreas made about downloading gobs  
of data is a huge concern for me on this right now.

>>> 2) We should move Adium to mtn.
>>
>> *ANYTHING* but mtn! Fetching an mtn repository (even just for read-
>> only access to compile the latest development version) involves
>> fetching a database file in the hundreds of MB (that's for pidgin,
>> probably more for Adium).
>> I've been using mtn for GSoC at libpurple, it feels like working back
>> in the 1980ies. Syncing changes takes about half an hour over my
>> 2MBit/s Internet connection (depending on the amount of changes on
>> both sides, but that was a usual number over the summer), where it
>> takes 5secs for svn.
>> Sean blamed that on the slow server, but why does a VCS require so
>> much computing power?
>
> Hm, it would be interesting to do an import and see how long that
> takes. Mozilla did that and hg was the only one with reasonable import
> times.
>

What's required for this?

>>> Moving Adium to monotone is, at this point, the path of least
>>> resistance. The people that we collaborate the most with, pidgin,  
>>> are
>>> using monotone.
>>
>> They moved far too early in the game, the other tools have become
>> much better since then.
>
> Any suggestions? git sounds like it's going to be be better "soon,"
> but I dunno.
>

Git is not acceptable. I've talked to Graham about that at length in  
person, it seems that they require some really weird stuff for you to  
get commit access, like local shell on the box. That'd make handing  
out commit bits very.. well, it would suck.



>>> Counter: They are not dead set on it, and the land of DVCS is
>>> changing
>>> rapidly (bzr is getting faster, git is getting easier to use, etc).
>>> Maybe now another VCS makes sense for both them and for us. Rlaager
>>> said that they might be willing to switch if it'll make life easier
>>> for them and us.
>>
>> Over the summer, I had to use darcs to fetch another project's
>> repository (right now, every project seems to have their own DVCS
>> program). This was just as simple to use as svn, and just as fast.
>> This was read-only though, I don't know how good the developer side
>> of it is.
>
> Is darcs fast enough?
>

Let's find out. Do you have time to make a feature grid/suckage level  
grid against all the popular vcs solution? Even the non-distributed  
kind?

> I really wish the adium-devl list wasn't fucking with the headers so
> we could keep the pidgin guys on the thread :\


What in the world? If there's a problem, ping Evan and cc me, explain  
what it is. First I've heard of this, mail works fine in Mail, I  
blame them :P

Chris