On Friday 03 January 2003 02:23, Peter B. West wrote:
> I'm deeply conservative about critical software like the repository.
> However, when (and if) the Apache repository goes to SubVersion, FOP may
> be obliged to go with it.
AFAIU there will be a choice, the CVS and SubVersion repositories
will run in parallel for quite some time.

>  If there is a choice, I would be strongly in
> favour of staying with CVS for at least six months after other projects
> had gone over, so that they can suffer the pain.
I consider 6 month after the repository is up "soon enough", although I
guess there will be an extended test time to check the operational
procedures (boot, backup, restore, failure modes), which may be part of
this 6 months.

> I would say that "ready for production" means that significant codebases
> have been maintained in SubVersion for long enough for the wrinkles to
> have appeared and been ironed out.
Well, SubVersion itself runs on SubVersion for some time now. I suppose
tigris.org will migrate right after 1.0 is out.

> How long was CVS around before gaining the level of acceptance it now
> enjoys?
CVS started as a directory tree oriented system on top of RCS with
a locally mounted filesystem as a repository. The network client/server
stuff was grafted on later, in response to popular demand and in
coevolution with the idea of public repositories accessible on the
internet. IIRC this started in late 1996 and it took well into 1998 to
squish major bugs and quirks and included some incompatible
network protocol changes. During this time it was already used
for major projects, according to the cvs lists. SourceForge started
somewhen in early 1999 with public CVS repositories.

SubVersion was designed as c/s system right from the start, mostly
in response to fix the deficiencies of CVS, in particular
- truly atomic commits (you don't want to know what happens if
  your network or, worse, the server goes down during a lengthy CVS
  commit)
- accessible through firewalls
- moving controlled files across directories
The real problem isn't that SubVersion could fail in situations where
CVS still works. The problem is the availability of client side tools,
specifically of GUI tools. That's not only an inconvenience for users
uncomfortable with command line applications, also only with the
availability of third party client side tools problems with protocol
conformance will be detected.

J.Pietschmann

---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]

Reply via email to