[RFC] making the null revision become real

[Aaron Bentley]

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

I like this idea in general.

Denys Duchier wrote:
> I'd like to propose to turn the null revision into an actual revision.  It
> should have revid "null:" i.e. bzrlib.revision.NULL_REVISION, and it should be
> present as the first revision in every branch's history.  

I think you also were including the idea that the revision would have
real metadata associated with it, e.g. a log message.  I'm not quite as
certain that this is a good idea.  For example, I think the inventory
format pretty much demands that there be a root id, and the null
revision is defined to have no root id.

The other issue is that, until stores perform escaping on the ids,
including ':' in the ID is incompatible with win32.  If we want to move
faster than that, perhaps we should change the string, since so far, it
hasn't been committed to disk.

> This means that revnos
> no longer have to be offset; revno=0 is simply the null revision.  It also
> becomes trivial to branch off a newly created empty branch.  And we should no
> longer accept a revid=None to mean the null revision.  revid=None should only
> mean that a revid has not been provided.
These all sound great.

Aaron
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.1 (GNU/Linux)
Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org

iD8DBQFD86660F+nu1YWqI0RAi9iAJ9ycNzsXFj7aKFPRsOENtejhEasjwCeOb9B
/PZljdH1S3tZBgNLHZgILZc=
=m7ZS
-----END PGP SIGNATURE-----