Is "bzr push" safe?

[Stephen J. Turnbull]

Barry Warsaw writes:

 > I suspect that it's more that most projects have a strong sense of
 > their main line of development, i.e. the "blessed" trunk.

Which really is another way of saying what I did, because VCSes aren't
used by *projects*, they're used by *developers*.  So developers have
a strong incentive to concentrate on producing patches for the
"blessed" trunk, and avoid "wasting" time on anything that doesn't
advance their current patch.

I can't do that.  I've never seen a codebase that wasn't riddled with
defects that don't manifest as runtime errors (missing or
unintelligible or style-guide-violating docstrings, obsolete comments,
etc).  So my private work has no blessed mainline, it doesn't go
straight forward.

Only git really allows me to be me, and "just commit" "early and
often" (like Richard Daley's advice about voting!), and fix up the
conceptual dependencies (DAG parentage) later.  mq, pipelines, looms
all provide similar functionality, but they require me to *think*
about what I'm doing with the VCS while developing, rather than simply
following the "early and often" rule, and worrying about presentation
to 3rd parties later.

You might think I'd prefer Darcs, which manages this stuff more or
less automatically.  There are two problems: "more or less" in
practice is "less" when it matters, and private history does matter to
me, though not as much as public history (which IMHO *should* be
inviolable *and indelible*, damn the RIAA and BSA anyway!)