Re: Flatten sysdeps/unix/bsd/bsd4.4 into sysdeps/unix/bsd

[Roland McGrath <roland at hack dot frob dot com>]

> and not (or rarely):
> 
> * Cleanup X would be better.  [Without explicitly saying whether it's OK 
> as is.]

Certainly one should be explicit.  There's not even a rare need to be
vague about what one means.  (Of course, people are bound to be sloppy
from time to time and sometimes a review with quick turnaround and
less perfectly clear communication is better than a more thorough
and carefully-worded one that takes longer to arrive.)

> * You should do this general infrastructural cleanup first.
> 
> * This needs testing on platform X.

Sometimes these are the correct answer.  Perhaps it's "rarely".
But it's a false economy to say everything must always happen
quickly regardless of the true degree of urgency for a change
(or lack thereof) or the long-term maintenance cost of doing
something hastily and poorly rather than deliberately with care.

We've made a huge amount of progress in a short span of time, and most
things have gone the way you wanted them to.  Frankly, I'm getting
pretty tired of having you and Dave jump down my throat every time I
say something can wait a few more days or even a week or two while
we figure out how to do it safely and correctly and give the people
who know the most about it some time for reflection.

We are redressing years of backlog while keeping things working on a
daily basis in a system that has a lot of subtleties to be considered.
It is going to take some time before it is either practical or wise to
operate in all the ways that will make sense in the well-oiled state
we all want to get to.  Your constant impatience is not helping to
achieve or maintain the comity that is the whole point of the exercise.

It's good that you are eager to do lots of clean-up work.  It's also
good that I am pushing back on false urgency for things with little
material gain and unknown risks that require careful consideration.
But every time we have one of these conversations, it's time that I'm
not spending on the backlog of patch review that addresses issues of
more true urgency and more significant and more immediate material gain.

I am immensely grateful for your work on libc and especially for the
devotion of time and effort that has enabled your very quick growth
into a clear leadership role in the project.  With each passing day,
it is harder and harder for me to infer that you have any appreciation
for my role.


Thanks,
Roland