[p4] Git/ BitKeeper comparisons

[David Ferguson]

Thanks to David (and Chuck and Tony and Robert and anyone else I missed) for
all your comments.
The scenario described has been recognized by support and a feature
enhancement (I think I have the words correct) is being planned for an
upcoming release.  I haven't tried it on 2006.1 but was informed that it
would not be available in it.  Yes, the code really is duplicated, not just
reported as integrated twice.  The more I think about the issue the more
complicated I understand the solution would be so I'm sympathetic to
support...
 But that doesn't stop a few of my engineers from being completely
frustrated...

Thanks for the info on Bitkeeper/git/distributed sources  It provides me with
a great deal of useful info.

-daf
 

-----Original Message-----
From: Weintraub, David [mailto:david.weintraub at bofasecurities.com] 
Sent: Friday, October 27, 2006 6:35 AM
To: David Ferguson; perforce-user at perforce.com
Subject: RE: [p4] Git/ BitKeeper comparisons

Here's an interesting take on Bitkeeper's distributed source archive
system: 

<http://web.mit.edu/ghudson/thoughts/bitkeeper.whynot>

While this is specifically aimed at BitKeeper, It can also apply to Git.

A lot of people use Git and BitKeeper just because Linus does, but that's not
usually a good idea. Distributed source systems have certain
advantages: You have no dependence on a single server and people on the top
of the hierarchy have total control without restricting source code access at
all. Anyone can checkout the entire Linux source, check it in and out to
their heart's content, and even pass the changes directly to Linus who will
simply push them into /dev/null.

They also have quite a few problems: Source code control is almost impossible
since each user can have their own complete copy of the source archive. And,
all integration is completely dependent upon a single gatekeeper.

I have never seen a corporate environment which was happy with a distributed
source system. The very few I've seen that use BitKeeper do not use its
distributed source abilities. Also remember that most OpenSource projects
have rejected Linus' distributed source model. Most either use CVS or
Subversion. 

Besides, what you are complaining about isn't a feature or bug due to
Perforce's central repository model. I can't say what happened. But, it could
be that change #45 was not reintegrated again, but only that Perforce reports
that the last integration took revisions 41 to 50 The Integrate database has
a "begin revision" field and an "end revision"
field for the "To File". Therefore, it can't report that it skipped revision
#45 because there's no skip revisions fields in the database record.

You stated that change #45 *seems* to be applied twice. Can you tell if it
was?

-----Original Message-----
From: perforce-user-bounces at perforce.com
[mailto:perforce-user-bounces at perforce.com] On Behalf Of David Ferguson
Sent: Friday, October 27, 2006 3:44 AM
To: perforce-user at perforce.com
Subject: [p4] Git/ BitKeeper comparisons

Okay ...Anybody want to summarize experiences of usefulness/effectiveness of
Git or BitKeeper for large scale installations -- We've been wrestling with a
particular feature/bug of Perforce for awhile now ...
Basically
BranchA has changes 1-50
BranchB integrates 1-40 (of A) into B
BranchB integrates change 45 explicitly into B BranchB then integrates 41-50
of A into B
 
Net result is that change 45 seems to be reapplied into B.
 
As a result of the above result, I've got a developer who is absolutely
convinced that we should shift to Git, or at least BitKeeper since they're so
much better at distributed source models.
 
I'll be handling a rebuttal concerning how good/ bad Perforce really is at
distributed source models (and the obvious corollary of how our code is NOT
distributable) but didn't really feel like spending a week playing with Git
and BitKeeper to learn their weaknesses.  So I figure I'll ask around for
Cliff's Notes.  Anybody got any?
 
-daf
David Ferguson
Manager of SCM team -- VMware
_______________________________________________
perforce-user mailing list  -  perforce-user at perforce.com
http://maillist.perforce.com/mailman/listinfo/perforce-user