[Mailman-Developers] Re: CVS branches?

Marc MERLIN marc_news@vasoftware.com
Sun, 21 Apr 2002 09:50:02 -0700


[moved to mailman-developers]

On Sat, Apr 20, 2002 at 05:52:30PM -0700, Ellen Spertus wrote:
> I could use some advice here.  I took a snapshot because I'm making a lot
> of my own changes to the code, i.e., adding dynamic sublists.  My

Yeah, I know about that problem :-)

> short-term goal is to do a user test real soon.  My long term goal is for
> the changes to be incorporated into Mailman.  

Barry and I  had a discussion by mail about  this recently. Basically he has
his  day job,  his  real life  [tm]  and he  pretty  much maintains  mailman
by  himself, which  in  other  words means  he  sometimes/often has  limited
bandwidth.
On one  side, reviewing other  people's code  for correctness, and  to check
whether they  really merge right with  his grand scheme of  things wrt where
the code is going to go, takes time.
On the  other side,  he's trying  hard to  get mm  2.1 out,  which obviously
should be a higher priority than adding code at this time.

Either way, some people will get a bit disappointed, you can't fully win :-)

Once 2.1 gets  out, we can all  release our own patches against  it, but I'm
hoping it's not going  to be like the linux kernel where  people now have to
deal with applying  multiple patches from outside the  main distro, possibly
with conflicts, or  with bugs that they  may report as bugs  in mailman when
the bug is really a bug with the patch

Ideally patches  that would  be of  interest to  many would  go in  the main
distribution, but  I don't  think that much  more new code  is going  to hit
mailman before the 2.2 cvs branch opens.
I've  however asked  Barry  if he  would consider  code  that is  reasonably
confined (either in the code, or because it can be globally enabled/disabled
in Defaults.py). To be  fair my mail was much longer  than that and probably
requires  some heavier  thinking, making  a policy  decision, and  setting a
precedent, so he's still thinking about it :-)

My guess is that really specific code  that wouldn't get used by many people
should probably  remain as patches,  which won't be  as big a  problem since
less people will  have to apply that  code, and you're also  lesss likely to
have to apply two patches like that and work out the conflicts.
What makes a patch fit for  inclusion in mainstream, or too specific, that's
left for the maintainer to decide :-)

> It seems my choices are:
> - working from a snapshot (what I've been doing)
> - getting my own CVS branch
 
I've been  working from CVS  and moving my patches  from time to  time, even
though it's time consuming, because many bugs got fixed in CVS
 
> Would you be willing to create a CVS branch for me (call it "mills") that
> I can write my changes to, merging in fixes from the main branch?  (I am 
> NOT asking for write permission to the main branch.)  Or perhaps you could 
> suggest a better mechanism.

That's a call for Barry to make, I'm just a user and contributer, but I have
no CVS access (well, none that I'm supposed to have at least :-D)
I  don't  know how  much  branches  help if  they  diverge,  but if  it's  a
temoparary branch that merges eventually, that could work.
Again, Barry's call, and remember that most of his energy is into getting mm
2.1 out :-)

Cheers,
Marc
-- 
Microsoft is to operating systems & security ....
                                      .... what McDonalds is to gourmet cooking
  
Home page: http://marc.merlins.org/   |   Finger marc_f@merlins.org for PGP key