[lug] Re: CVS and Subversion

[Tom Tromey]

>>>>> "Jani" == Jani Averbach <jaa at jaa.iki.fi> writes:

>> I'll list them off in some subsequent post. The upshot is that SVN
>> won't scale well to the enterprise level no matter how much snake
>> oil CollabNet pours on it."

Jani> If you make this kind of statements or forward them, I would like to
Jani> hear your opinion about following systems, please:

Jani> GNU Compiler Collection
Jani> http://gcc.gnu.org/svn.html

Jani> I am especially interested why do you think these are not "enterprise
Jani> level" systems, and what kind of scalability issues there are
Jani> currently (there are some, for sure with projects of this magnitude.)

FWIW I work on GCC.  We switched from cvs to svn sometime in the last
year, I forget exactly when.

Overall svn is superior.  For some common operations it is noticeably
faster.  It uses more disk space for a working tree (due to the text
base) but this means that plain 'svn diff' is very fast, and 'svn revert'
is a local operation.  Also 'svn status' is local which is very
handy (with cvs I had to have a whole suite of helper scripts to deal
with this).

I didn't do the migration but I gather it went fairly well, even
though GCC's repository is big and ancient and had all kinds of
ill-advised cvs hacks in it (evil stufff like rebranching by moving
the branch tags).

As I recall svn was initially found to be too slow for a couple
operations (svn annotate, I think), but Dan Berlin (who works on both
svn and gcc) fixed these problems.

As a user the migration has been easy.  svn and cvs are pretty similar
from a command-line point of view.

Tom