[rescue] DEC keyboards

[der Mouse]

>> For example, the SBus code is outside arch/ so it's usually thought
>> of as MI, even though it never gets used on anything but a SPARC and
>> thus really is MD code in some senses.
> Except I think there's probably nothing stopping someone from
> building a PowerPC-based (for example) SBus machine.

True as far as it goes.  If that were to happen, the code in dev/sbus/
would probably make large strides towards machine independence.  (The
MI/MD split is not a totally black-and-white thing; it's a matter of
degree in many respects.  For example, code could be basically MI
except that it breaks badly on big-endian arches.  Or on 64-bit arches.
Or those with separate address spaces for kernel and user.  Or....

> Would it be incompatible with all of NetBSD's SBus drivers, probably,
> but that could be rectified fairly easily.

Well, maybe.  Depends on a lot of factors, only some of which I know.

>>> I could never stick with a release that old though.
>> I'm not saying anyone else should.  But that's when it became
>> unignorably clear NetBSD was going in a direction I didn't want my
>> own playing-with machines to go, so that's when I stopped following.
> In what regard?  Bloat?

Not really, except to the extent that code I never use and don't really
want qualifies as bloat simply because of those characteristics.
(Which may be a defensible position, but not one I want to take.)

More a matter of emphasis.  I saw it as a shift from being a research
and development system to being a production system (this wording is
because it could very well be that NetBSD never actually changed its
focus, and I had instead misread its focus).  To pick the specific
example that tore it for me, when they split up /etc/rc, they eased
mechanized sysadminning at the price of complicating manual
sysadminning.  (Split-up rc does ease some manual tasks, but only for
tasks it specifically anticipates.  Sort of a "when we want you to
tinker with that we'll give you a setting for it" attitude.)

I run my home machines because I enjoy tinkering with them.  Modern
NetBSD is not that much fun to tinker with any longer; there are too
many layers of "we're going to give you this slick-looking UI to do the
things we think you should want to do; don't bother your pretty little
head about what's going on under the hood, because we know better than
you how you should run your system".  Since the sort of setup they're
optimizing for is relatively common, it's still a fine system for lots
of applications - indeed, at $DAYJOB we run modern NetBSD on most of
our infrastructure machines, and I think it's a right choice.

But tinkering by someone like me is something they will cheerfully
sacrifice to be better suited to someone who just wants a turnkey box
to run DNS - or Apache - or MySQL - or whatever - on.  And they've done
so, repeatedly, to the point where the inappropriateness of sticking
with a 6.5-year-old system is less than the inappropriateness of trying
to continue doing what I so love on -current.

/~\ The ASCII				der Mouse
\ / Ribbon Campaign
 X  Against HTML	       mouse at rodents.montreal.qc.ca
/ \ Email!	     7D C8 61 52 5D E7 2D 39  4E F1 31 3E E8 B3 27 4B