[Gnu-arch-users] Re: Improving Archive Cached Revisions

[Andre Kuehne]

Matthieu Moy wrote:
> If you search for "delta compression" or if my memory's good, "summary
> delta" on this list, you'll find some discussions on the topic, yes.

Okay after going through the two recommended threads, i'd like to say
that Aaron's proposal would let me do pretty much what i had in mind.

Regarding the other (partly very complex) caching schemes (which are
meant to work mostly without user intervention): I doubt one of them
is optimal in every way (disk, cpu, bandwidth usage, number of
transmitted changesets, whatever).

Every project is different in disk quota, project activity and size,
number of archive users, archive location and availability. And
archive maintainers may differ in what goods they value most.

Therefore i really like the way archive cached revs work: You can put
them where you want, when you want. They just should be cheaper in the
sense of disk usage. So i don't think arbitrary deltas (and i think of
them as enhanced archive cached revs) are too much flexibility from a
users POV.

The analogy to backups is indeed obvious. There, you have full,
incremental and differential backups. You can combine them in any way
to meet your needs. If this is too complex for anyone, he can use some
standard scheme or program.

Until now, tla provides basic operations and leaves it to third party
tools to provide more high level services.  I would consider adding
the ability of creating arbitrary deltas very much in this tradition.

Best Regards

-- 
     _   _
_/\_| |_( )_
Andre Kuehne