packs and knits

[Aaron Bentley]

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

Robert Collins wrote:
>> (The fulltexts can be derived from the deltas, so
>> they are *derivable data* that could optionally be stripped.) 
> 
> Equally, the deltas are derivable from the fulltexts ;).

Oh, you're proposing a VCS where all revisions are stored as fulltexts?
 Pleased to meet you, Mr. Torvalds. :-D

>>  That
>> would make generation of bundles faster, since it would be more a matter
>> of filtering out fulltexts than generating anything new.
> 
> Presuming that the deltas are mpdiff deltas?

No, I'd be open to shipping another delta format in bundle format 5,
assuming it was efficient.  Or we could mix knit deltas and mpdiffs if
necessary.  Or something.

>> No, but 0.19 is expected to.
> 
> Ok. Is it going to be traumatic if I update the current pack format
> then, as it hasn't been released? That would be a little simpler. I can
> do it today.

I don't think it would be traumatic.  But I think it's a bit risky to do
just before release.  If you decide to proceed anyway, you may run into
bizarre errors in get_invalid_bundle, because it's sensitive to the
serialization format.  I can help out if you do.

Aaron
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.6 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFGuPbN0F+nu1YWqI0RAry1AJ0W9XDBp5LZBWR3SR19HdAHKNWV2ACfY9Lz
p0levAU6GiFIsx8fEj324gU=
=waGM
-----END PGP SIGNATURE-----