Patch distribution and committing

[Rodent of Unusual Size <Ke...@Golux.Com>]

The ongoing discussions about how code moves from proposed patch
to committed have been in a couple of inappropriately named
threads, so I'm starting yet another one with a hopefully better
subject line.

There are actually two separate discussions going on.  One has to
do with whether our voting procedures are too cumbersome, and the
other regards whether it's better to distribute proposed patches
through mail or through CVS.  These two seem quite related to me.

Four vocal people have weighed in on these; I would really like
to know the opinions of people who *aren't* vocal, *don't* vote,
but *do* track the CVS tree through cvsup or some other mechanism.
Particularly on the subject of whether patches in a CVS directory
would complicate or simplify things for them.

For myself, I'm willing to try any of these different alternatives,
both for voting procedures and for patch distribution.  This is the
only distributed development project in which I've ever been involved,
so I really don't have a valid opinion.  I *predict* that distributing
patches through CVS would simplify things for *me*, but I could be
wrong.  I have no predictions about the voting alternatives.

Let's not be too conservative, eh?  Let's try new (or different) things.
The people involved aren't quite the same, the environment isn't
quite the same, and the proposals aren't the same.  If there isn't
a concensus, I think it behooves us to gather experimental evidence
rather than assume we know all the answers - particularly since
disagreement indicates that the answers aren't the same for everyone.

#ken	P-)}

Re: Patch distribution and committing

[Brian Behlendorf <br...@organic.com>]

Okay, I'll repeat a proposal I made in another thread here just for
consistancy's sake.

At 05:12 PM 1/8/98 -0500, Rodent of Unusual Size wrote:
>The ongoing discussions about how code moves from proposed patch
>to committed have been in a couple of inappropriately named
>threads, so I'm starting yet another one with a hopefully better
>subject line.
>
>There are actually two separate discussions going on.  One has to
>do with whether our voting procedures are too cumbersome, 

I think they are.  This summer we operated under a "lazy voting" system,
whereby those with commit access could implement changes since we were in a
highly experimental time frame, and if a big mistake was made it was easy
to rollback.  Now we're "supposed" to be in feature freeze, but really
worthwhile new features keep coming along.  I think so long as we all
continue to respect each other's opinions, and a patch author has the
humility to roll back a patch when a change made doesn't meet with
approval, and that anything really major or questionable is passed by the
group before committing, we can safely allow individuals to commit.  We
should develop a procedure to revoke commit privileges, though, in case
anyone becomes too "renegade".

>and the
>other regards whether it's better to distribute proposed patches
>through mail or through CVS. 

My proposal, as described in my last message, is to put patches in a
subdirectory on dev.apache.org automatically whenever a message to the list
with the subject of [PATCH] appears (not Re: [PATCH] of course).  Then
we'll have an easy URL we can reference a patch to, and testing those
patches won't have to involve digging through monthly archives of new-httpd.

	Brian


--=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=--
specialization is for insects				  brian@organic.com