plain/rich/subtree differences

[Aaron Bentley]

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

Robert Collins wrote:
> Hi Aaron, as you guessed I am writing a new serialiser. I'm not aiming
> for composition or other things yet, in this iteration I'm aiming to
> just finish making commit scale well.

Cool.  It'll come sooner or later.  But when an idea catches flame,
better to share it.  At least now I know that we probably want
bidirectional inventory deltas.

> I'd rather not require everyone move to subtrees, nor write 3
> serialisers, so I'm writing one that can be parameterised.

Okay.  I'll just mention that the advantage of having different formats
is that we get hard errors if someone tries to move data into a format
that doesn't support that model.  But of course there are other ways of
achieving that.

> heres what I understand the differences to be:
> subtree:
>   / is versioned and has a fileid that may be TREE_ROOT
>   paths other than / can be 'tree-reference' kinds
> rich-roots:
>   / is versioned has has a fileid that may be TREE_ROOT
> plain:
>   / has no last-modified revision and must have the id TREE_ROOT

Confirm.

Aaron

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

iD8DBQFHfU6L0F+nu1YWqI0RAiOVAJ9xc1krwnKev68TwD4mGUQTr7SVagCeJ845
/V0Zbncf/pE3jpKpKOuuCJE=
=RmTi
-----END PGP SIGNATURE-----