backward diffs in knits?

[Aaron Bentley]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Matthieu Moy wrote:
> Aaron Bentley <aaron.bentley at utoronto.ca> writes:
> 
>> Matthieu Moy wrote:
>> [...]
>> But if you look at it, I think you'll find the advantage is all due to
>> revisions 6+.  If we pretend the changes stop at 5:
> 
> You're also changing the position of snapshots. Take 20 revisions
> instead of 15 to keep snapshots at the same position, and you'll get
> the benefit for each non-snapshot revision.

I think this is basically the same thing as I was saying: "The problem
with my example is not that it selects only a small change, it's that it
assumes an unrealistically short history."

> In a perfect world, the commands should be able to deal with a mix
> "normal branch" + bundle at the same URL.

Sure.  I think first we need faster bundles, and then "branch bundle"
will make sense.  Maybe it's worth supporting "branch from tarfile" too.
 (where tarfile is a tarred-up branch).

Aaron
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.3 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFGHBQf0F+nu1YWqI0RAlqiAJ9woFQR5nSOFj24Zv28x/x7DAYQbgCcDcR9
GIU2NR6FlhGClikBZNqFLiI=
=slTk
-----END PGP SIGNATURE-----