bzr-svn svn-import and history across svn "renames"

[John Arbash Meinel]

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

Chris Hecker wrote:
> 
>> Rename inference would have severe performance consequences,
>> primarily because bzr-svn supports more than just one-time imports.
> 
> Sure, it just seems like giving the user the option even if it's slower
> is better than no option that doesn't lose information.
> 

The problem for bzr-svn is that it uses deterministic revision ids
(which is how it is able to interoperate with svn across multiple
clients). However as soon as one of us uses a different flag to convert
the repository, our repos are actually incompatible. Even worse if you
keep deterministic revids then you have potential corruption (because I
say I have revision 'foo' and so do you, but they point to different data.)

You can have the flags as part of the revision-id, but that also forces
2 people to no longer interoperate. (Not to mention what happens when
you push that back into svn...)

As such, using an svn => fastimport stream and then using 'bzr
fastimport' is significantly better whenever you can have a "canonical
conversion". It lets you do any pruning, or optional work in that
conversion, and everyone who wants to use bzr just uses that copy of the
conversion.

John
=:->

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (Cygwin)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/

iEYEARECAAYFAkwreb0ACgkQJdeBCYSNAANQAgCbB0Epi8/HnX5EQbK/36nLXnVH
PNAAn0I6gRg455W5hgs4SQWd3Vt9abfl
=XiRe
-----END PGP SIGNATURE-----