Chris Demetriou sez:

   Serial port problems have been around on the i386 port for a long
   time.  Given that, if the problem were considered 'important' by the
   people who maintain the i386 port, it should have been fixed a while
   ago, no?  Why take the chance that things will break even further --
   for the i386 and other ports -- by rushing something in right before a
   release, if it's been broken for so long?

This is a good point...

   It's not like people are saying: "gee, they're thinking about pushing
   out a new release, i bet they'll find a magical fix for the serial
   driver right before they do."

s/bet/hope/p, and yes they do :-).

   _My_ concern, as the Alpha port maintainer, is for _my_ users.  I
   don't want _my_ users to be screwed by hasty last-minute changes to
   fix a long-standing and, apparently from the amount of work that's
   historically been done on it, low-priority bug in another port.

This almost looks like an argument to make the COM driver platform-dependent,
or at least #ifdef'd (I know, I know -- YUCK!).  One could also argue that
one port should not need to suffer because of the shortcomings of another's
platform.  I mean, you explicitly stated that the Alphas don't have variable
hardware interrupts -- everything interrupts at the same level.  I gather,
then, that the software needs to jump through hoops in order to make the
Right Things [TM] happen.  It seems to me that taking the lowest common
denominator in this case is not the best idea, and perhaps the dependencies
and/or independencies need to be re-evaluated.

I'm sorry if it looks like I'm being an ignorant cretin and blowing smoke,
and there may be more here than is meeting my eye, but I get this hunch
I'm not entirely off the mark, here.  Elucidation on any points I am not
grasping would be appreciated.

				--*greywolf;
--
Sun could have remained quite profitable by staying with BSD-based OSs, and
they wouldn't have pissed off NEARLY as many customers.