Bazaar still below the radar when evaluating VCS tools

[John Arbash Meinel]

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

Stephen J. Turnbull wrote:
> Matthew D. Fuller writes:
> 
>  > Well, sure.  You can do anything by creating a new history and
>  > browbeating people into considering it the authoritative one  8-}
> 
> It's not about creating new history, though.  It's about presenting
> the old one in a way which reveals the big truth that the observers
> want to know, rather than a slew of small truths that nobody really
> cares about (until the investigative reporter from National TV shows
> up, and uses them to raise her ratings along with the ad revenue of
> her station, and in the process convince the Budget Office that you're
> all a bunch of wankers and you lose your funding, of course).
> 
> It *is* nice to have a bisectable history.  There are two ways to get
> it.  One is to never commit until you are sure you have a sufficiently
> clean version.  The other is to commit when you feel like
> checkpointing, then edit out the commits that don't say anything
> interesting, or are broken.  (I think of this as a sort of "persistent
> undo" function of my editing system, rather than as part of the VCS,
> even though it's implemented via the VCS.)

Or (3) commit what you want, but only land things into your 'stable'
branch when they pass the test suite. (preserving the lefthand history
as all of the lefthand parents are 'bisectable')

This is, in fact, bzr's development structure. You aren't guaranteed to
be able to bisect any particular branch but:

1) you can bisect the mainline
2) you don't have to worry (as a developer) about cleaning up your
history just to get something landed.

John
=:->

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

iEYEARECAAYFAkuG4kEACgkQJdeBCYSNAAPAswCfaB9Qvq7na4ag8eCeJKbImwtF
ecUAoKYU/L2h56LrfBIbIAxPgr1u79HU
=sDM8
-----END PGP SIGNATURE-----