Re: [tlug] bash and grep and diff

["Stephen J. Turnbull" <stephen@example.com>]

Pier Fumagalli writes:

 > In all the realities I've had the luck (or misfortune) to observe,
 > code integration and review happened way more often where VCSs have
 > a central repository, while in fully distributed environments,
 > developers tend to more easily go off on their own tangent for a
 > few days/weeks/months and integrate less often.

This is a problem in Linux?  X.org?  I think not.  In fact, in my
(limited) experience, it was a bigger problem with CVS.  Not entirely
because of the central repo, but because most wannabe contributors
sent us absolute crap, we said "no" a lot, we had to control what went
into it carefully.  In many cases, without a repo of their own, they
became discouraged and never tried again.  I *think* (but don't have
numbers) that since we switched to hg, a lot of people have become
more active because it's much easier for them to integrate our changes
and see a path to eventual contribution in the future.  I can't say
the first submissions have improved much, though. :-)

 > And it's true that even with distributed repositories one can set
 > up a master one and do all the integration in one single place, but
 > developers have the tendency to do "git commit" more often than
 > "git push"...

But this is a bogus comparison.  I've experimented with git workflows
where I do git commit every 300 keystrokes (painful) and with every
file save operation (works OK, but could be improved with more
tools).  It is reasonable to git commit at least with every save,
although probably not a good workflow for most people.  There is no
way you want people pushing that often.

The real question is "do you git push as often as you cvs/svn commit?"
In my case, the answer is sometime yes, sometimes no.  With git
(actually, XEmacs uses hg) I push trivial changes and obvious fixes
much more frequently.  I tend to let big changes season more, get the
docs and tests written for example, before pushing.  IOW, in my
experience, DVCS allows the *timing* of a push to be more accurate.

 > It's true, all based on developer's culture and habit, but as
 > everywhere else "freedom" and "control" stand at the opposite ends
 > of the scale...

Not at all.  True freedom is enabled by control.  Any baseball pitcher
or evangelical Christian will tell you that (although they have
different ideas of who should be in control :-).  My own opinion is
yet otherwise....  And good managers will tell you that the control
they *need* is greatly enhanced by giving their people as much freedom
as they can when they don't need control.