(More opinions of mine...  8-)

> I think that the divergence is on the issue of what exactly "the right
> thing" is.  There are two perspectives on this: the users, and the
> developers.

Unfortunately, these two perspectives are _not_ so clear-cut.

It is my opinion that, in the long term, both the users _and_ the
developers benefit most from a well-designed and carefully implemented
kernel.  (Certainly, short term benefits for users, and perhaps even
for the devlopers, lie in integrating things and coding things more
quickly.)

If things are designed, implemented, and integrated carefully,
developers will benefit in the long term, because the kernel they're
working on will end up containing less 'hair.'  This eases
development and debugging, and makes for a more stable and reliable
system.

"Users" also benefit in the long term from developers' care, because,
while they may not get the latest driver immediately, when they do get
a driver it'll be a good one.  They'll also benefit from whatever
stability good design confers.  (You'll note that i also think
multi-platform support also benefits users for exactly the same
reasons: it encourages solid design.)


> For NetBSD, essentially
> whatever works for core developers is the right thing.

This is _not_ the case.  Yes, if a piece of software, or a decision,
comes all the way to 'core,' then it's 'core' that'll pass judgement
on it.  but, especially for port-specific features, drivers, etc.,
things are left entirely in the hands of port maintainers.

In my opinion, the things that 'core' actually ends up being
responsible for the following:
	(1) designing 'big' things, then implementing them,
	(2) making sure that things go relatively smoothly, and that
		nobody does anything 'insane' to the system,
	(3) keeping track of who's doing things, who wants to do
		things, etc.
	(4) a whole lot of administrative garbage (some of which
		are mentioned in (2) and (3)).

It's not a glamourous job, or anything...

'core' also holds the 'keys' that determine who gets access to the
source tree, though we try to give people who have reason access to
the source tree, if we know that they generate good code and/or they
have a good reason to have access.  (There are a few exceptions to
this general statement, which should not be discussed here.)


> Probably
> some years from now, once the core team has lost interest and energy,
> NetBSD will quietly become an intellectual curiosity along the lines
> of Nachos or Sprite, but not quite as cutting edge for it's time.

I could believe that.  Indeed, i think that Microsoft has virtually
guaranteed it.  (Hell, even 'OS programmers' at major research
institutions, are thinking of using things like NT as a base for their
work...  What does that say?)

Frankly, i think it's relatively unlikely that _any_ UNIX-like system
is ever going to be a 'large' portion' of the PC market.  However, I
can't worry about that now because i'd rather be hacking, and i've
already got enough garbage to think about (see "(4)" above 8-).



cgd