bzr bundle > file

[John Arbash Meinel]

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

Martin Pool wrote:
> On 8/8/07, Aaron Bentley <aaron.bentley at utoronto.ca> wrote:
>> Resurrecting the old "bundle" implementation wouldn't fix your use case,
>> because "bundle" always used the default format, and the default format
>> is 4.
> 
> (To try and summarize an irc conversation)
> 
> I don't care much about the old implementation; if it was ugly then
> I'm happy to get rid of it too.  However, I don't think we should just
> suddenly change the meaning of the command.  People are accustomed to
> writing 'bzr bundle ../OTHER > FILE" and that should still work.
> Ideally that can be just syntactic gloss on the current
> implementation.
> 

As we are getting rid of the command anyway, I think we should just
restore it to its former ugly glory and just deprecate it. Make it
generate the old format bundle, and make it have all the warts preserved.

I know Aaron feels very strongly about it, but I've used "bzr bundle
- -r-2..-1" for some very decent results. I realize it *can* generate a
bundle that cannot be applied. But I also know that "bzr send" will
always require connecting to the "remote" branch and doing a common
ancestry check. (Which has fortunately gotten a lot faster.)

Ultimately, I'm happy enough to migrate away, but I think as a public
command that a lot of people used, it needs to be properly deprecated.

John
=:->

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

iD8DBQFGuddfJdeBCYSNAAMRAj3pAJ9Sa7h28Kq9k2bacG9+D/abOKBd6gCgx+D1
FcbevIh7A3y3DaWCJa+kiww=
=5eWa
-----END PGP SIGNATURE-----