Re: How/where do we submit patches now?

Larry McVoy (lm@bitmover.com)
Sat, 06 Jun 1998 14:11:40 -0700


: Do you know if there was any discussion with Linus about formally
: spreading out this responcibility to other trusted people?
:
: I would like to do more for linux, but why put in the time if it's next to
: impossible to get the stuff even looked at? For instance the initrd-archive
: thingy I wrote I haven't even tried to fit in anything >.90's. I also wanted to
: clean some things in the rd code. These haven't been done because it's doubtful
: they'll ever see the light of day.
:
: With how big Linux is now, it needs to be better orginized.

Yes, there has been some discussion. Linus acknowledges that there is
a bit of a problem and is taking some steps to resolve it. The first
step was to start using Jitterbug as a patch queue, so at least things
don't get completely lost.

A longer term answer is also in the works. I've quit my job to focus
100% on a new distributed source management system that may help a great
deal; that remains to be seen. Linus & I and others have sat down and
discussed this at length and we are converging on a set of tools which
will help track things in a distributed manner.

Goals:

. provide a globally distributed source mgmt system
. stick the kernel under source control
. turn patches into version controlled things called "change sets".
. change sets can carry history (both deltas & comments) so that
can be used to fan out the load to Alan/David/etc.

Features:
. It's not CVS - CVS has one repository and multiple clear text
work areas. In my stuff, every work area is a repository and
has all the revisions. So you can work while you are disconnected.
. I revision control /everything/ - file contents, file names,
symbolic tags, etc. So when you update, renames propogate like
they should.
. Free for free software development.
. Works quite similarly to what people do now - you mail patches
around. The patches have more information in them, they include
the delta comments, tags, renames, etc. Thanks to Eric Raymond
for that idea.
. Change sets let source browsers ask questions like "this line of
code is weird - who put it in and what are all the other lines of
code in all the other files that went in as part of this change?"

Schedule:
. My best guess is a working prototype available for public use in
2-3 months. The base revision control is pretty close to completion
(I've been using it daily for about 6 months, I'm tieing up loose
ends); I need to redo the repository resync / resolve code in C
(it's currently in perl).

More info:
. Mail me, lm@bitmover.com, and/or watch

ftp://ftp.bitmover.com/bitkeeper

Help: . The most useful thing to me right now is people that have thought
about source mgmt over the years and can say stuff like "it had
better solve problem XYZ". I've discussed it with a lot of people
and it seems pretty good, but I like people that want to poke at it
and try and break the model. I'll have a majordomo list up on
bitmover in a week or so; stay tuned.

. If you are familiar with SCCS and want to test out the base revision
control, mail me. If you aren't familiar with SCCS and you are
freaking out because it isn't RCS, fear not - I give you "ci" and
"co" compatible interfaces to the system in addition to the SCCS
interfaces.

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu