On 1/19/2011 3:44 AM, Dan Creswell wrote:
Ah, so my measure of the merit of your suggestions includes "how many other
users want this/like it etc".
Unquestionably, some of the suggestions fall more into the realm of
preference than anything else; where that's true, I was hoping for some
initial feedback from the devs. Others potentially involve questions
of soundness/methodology, and I believe the discussion here has been
helpful in shedding light on those areas. Thanks again for your
contributions.
I certainly think this work would be best done in a separate interface as
I'm reasonably sure at this stage that there will be some implementation
surprises along the way that probably shouldn't be exposed on the "stable"
interfaces we already have.
In fact, I'm wondering why you wouldn't just code it up in a branch/scratch
for yourself and see what happens.....
Oh, that may happen.
I wanted a discussion first because:
1) The discussion would help expose gaps in my thinking. (And it has
done that: for instance, low-memory clients weren't on my radar because
they were never in the usage pattern at my company.)
2) Many of the suggestions can be implemented separately. If something
was overwhelmingly popular, there would be sense in doing that first.
If something was drastically wrongheaded, it would be best to not allow
it to sink a monolithic set of changes (especially where it involves
creating a parallel implementation of space).
3) If others were enthused about a particular suggestion, there might be
more people willing to contribute and/or check over the code.
The last question, of course, is my availability and how exactly to
split my time between such a branch and whatever other contributions I
may be able to make. I'll figure this out myself (naturally), but
means I won't be contributing a branch of all my proposed changes tomorrow.
jamesG
