Defining specific problems and handwaving at solutions (was Re: What's Canonical thinking about Bazaar?)

[John Arbash Meinel]

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


...

>> But I know that getting the code from the "works fine" point to the
>> "in bzr.dev" point is going to be a long, miserable, exhausting
>> process
> 
> Part of the problem is that "works fine" means different things to
> different people. Most people submitting patches care mostly about the
> "business value" of their code: it solves a bug or requirement and is
> therefore "good enough". Rightly or wrongly, stuff like API design and
> testability are less important to the majority of submitters. OTOH, the
> core contributors *must* care about these or the overall technical debt
> gets out of control.
> 
> The effect is predictable: only really simple patches, or those with a
> champion, get through in a timely manner. The base still evolves but not
> necessarily as rapidly as it could. The majority of innovation moves to
> the plugin layer.
> 


Patches need a champion, either the person doing the initial patch sees
it through to the end, or it transitions to someone else to do so. That
IMO is a fundamental part of software. *Somebody* needs to finish it.

It is possible that we should try to more-quickly adopt community
contributions and finish-them-off.

Having a Process that requires code-review and tests is a barrier to
entry. However, aside from some spectacular failures like stacked
branches, we've been almost regression free. We have a trunk that
everyone can be comfortable running day-to-day without worrying that
things will start breaking unintentionally.

I honestly don't think I would be comfortable running hg tip day-to-day,
rather than released versions. (You may start running it just before a
release to try and shake out remaining bugs, but compare that to
updating a couple times a day...)

So there are barriers, but there are also benefits. I don't think we are
at an optimal tradeoff yet, but I don't think the answer is to drop the
test suite, either.

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

iEYEARECAAYFAkr2BpgACgkQJdeBCYSNAAPq0wCeOVbP/2iFVM3/TfuMREkJUxiu
wN4AnizOumK/Ak5S9d8iSEDHFNmnmP0t
=cFMe
-----END PGP SIGNATURE-----