Re: [glob2-devel] mercurial switch

[Kai Antweiler]

> I agree with the mercurial switch.
>
> I don't think an cvs history import is really that nesseccarry. We can
> backup and maintain a copy using CVS, and get history from it. I
> personally don't use it allot except for recent changes, and I've
> never felt the need to do a diff with a 4 year old version of a file
> with the current version. As long as we maintain the old CVS, on a
> read only mode, the web interface on Savanahha can be used to look on
> old files.

To make things clear:
Importing the whole cvs MAIN branch works well and is no problem.
Importing the branches does not work for me.
I see no reason why we shouldn't include the history of the MAIN branch.
(Ok, the repository would be smaller.)


> The only real loss is recent changes and the ability to compare
> distance versions, which hasn't come in particularly usefull. While
> there are certainly "fundamental" problems that people stick by and
> having old copies is ohh so usefull just in case, from a practical
> point of view (and from the point of view from the current developer),
> they aren't really that usefull. I've never felt the need to use those
> kinds of features. Great in theory.

As we all know only part of the code and mangle part of it by
our bugfixes and new features, it is extremely valuable to be able
to look into the past.  When we look at the time some code was
implemented, we'll have a chance of finding out what it should do
and whether this can still work (or repaired) inspite of our edits.

Binary Search is very quick:  With 13 tries you can find what you're
looking for even with 4096 revisions.  Our MAIN branch currently
holds 2500 revisions.


> But in reality development will not be hindered by cutting off version
> and history information from Mercurial. The data is still available in
> CVS if its absolutely nesseccarry.

cvs is no good.  Using history in cvs is not as easy as in mercurial.
-- 
Kai Antweiler