'fetch'...

[John Arbash Meinel]

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

Robert Collins wrote:
| Other than ghost filling, is there any reason we can't change the usual
| fetch path to be
| target.insert_data_stream(source.get_data_stream) ?
|
| Thats roughly what pack to pack fetches are (though not expressed like
| that); it would make the variation between repositories in terms of
| transactional/non-transactional facilities easier to manage I think.
|
| For ghost filling, I think having the search select the ghosts will be
| sufficient, a SearchResult can list disjoint revisions.
|
| -Rob

We probably could. Though at the moment the pack-to-pack get_data_stream() uses
a very bad order. (It still sends the data as texts then inventories then
revisions, though the file should be stored as revisions then inventories then
texts.)

The other bit is that it would probably disallow optimizations like "if copying
all revisions from a pack, just copy the raw pack file" or even hardlinking if
possible.

Otherwise, it would be nice for the standard copy mechanism to be consistent.

John
=:->

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.5 (Cygwin)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFH+8p4JdeBCYSNAAMRAubDAJoD9lhb89oix7gKu79puMcy2w1i0ACff6gu
cBzHqA+jzLBF40p/MSHtb4Q=
=KSZn
-----END PGP SIGNATURE-----