I think the recent thread here started by someone who was told by
Introl employees that NetBSD/i386 was *only* usable  with ISA cards,
for machines with more than 16 Mbytes of memory (or whatever it was),
should be sufficient evidence for anyone that there's a widespread
public perception that NetBSD has problems on x86 maihines. That
may be a *mis*-perception.   It's still there.  And I'd hope Core
would like to do something about that...


Rick Kelly <rmk@rmkhome.rmkhome.com> writes:
>The VLB and PCI busses are there for a reason.  A Buslogic VLB controller
>costs less than a Adaptec 1542, and is a lot faster in the bargain.

uh, if you're willing to donate a Buslogic controller to everyone who
already *has* an Adaptec 1542 who might run NetBSD, and a new
motherboard to everyone who would need one, then I'd agree there's no
problem.

I'm asserting that a significant fraction of the "great unwashed" out
there see NetBSD as ``not real'', precisely because it doesn't support
systems with an Adaptec (or other bus-mastering ISA card) with more
than 16 Mbytes of memory, or that might one day have more than 16 of
memory.

As it is, there's an enormous advantage to being able to support
ISA bus-mastering DMA hardware with more than 16 Mbytes.  Any site
that has even *one* such machine has a strong reason to use
soemthing other than NetBSD, because they'll need to run something
else on that other machine anyway.  So why not run that "something"
on all their  machines?



And, tmm@mcci.com (Terry Moore) writes:

> I don't happen to use Adaptec SCSI boards. 

Lucky you.  Some of us do.


> should use FreeBSD, because NetBSD/i386 just isn't *REAL*, because to
>Please criticize if you like, but to say that NetBSD/i386 isn't real
>is a bunch of baloney.  If you want something else, go somewhere else,
>but don't waste our time with this garbage.

That leaves me wondering if Terry even read what I wrote....

I'm trying to say that NetBSD doesn't support certain x86 hardware
configurations, and that that is, in and of itself, enough reason for
a lot of potential users to *NEVER* run NetBSD, because it won't run
on *all* the hardware they'd like an OS for.  Whereas other OSes, like
FreeBSD and Linux, *will* run on all the hardware they want an OS for.
(Or at least, all the x86 hardware.)

I'm glad that NetBSD runs on all the systems you have.  What I'd like
is to be able to run a "clean" Unix on the hardware I have available,
and that lets me do all the things I currently do.  That includes x86
hardware with Adaptec ISA cards.  It also includes being able to
hot-swap PCMCIA cards.  The recent proposal to extend the static,
boot-time configuration to allow dynamic *removal* of devices deals
with, at best, one-third of the problem: hot-swapping a cards out.

Another 1/3rd is being able to insert devices into a booted system and
have them dynamically configured.  The last third is running
user-level scripts to do device-specific configuration (e.g.,
configure IP addresses on Ethernet devices and add routes) when a card
is hot-swapped in.

NetBSD/i386 just won't do that.  It won't run on a 486 with an Adaptec
1542 (which is probably about as fast as the Decstations I build
NetBSD kernels on.).  I think Terry is falling into a fallacy here:
NetBSD is a real alternative for him, so it's a real alternative for
everyone.

I think the best thing that the Core Team could do for NetBSD between
now and the 1.2 release, is to add *some* kind of support for ISA DMA
bounce-buffers.  Because, factually, NetBSD *does8 have a problem
there, and it's widely known that NetBSD has *some* kind of problem
with x86 machines, even if the exact details aren't always right.


And lastly, don't shoot the messenger.  If enough people tell me to
``go somewhere else'', I may just do that :).  I'm getting exasperated
with people saying I'm running FreeBSD [sic] as it is.