>> We should, yes. And someone (I forget off the top of my head) had 
>> integrated a CPU level emulator with a framework which passed all OS calls 
>> to the native CPU.
>
> That would be nifty!  That would really help bootstrap new architectures 
> (like parisc).

That would be.... I'll have to find the posts about that and check up on 
the status.

>> (50 pin half height SCSI drives are impossible to find with more than 9 
>> gigs of space).
>
> At reasonable prices, yes.  But larger SCA drives are available for US$20 
> nowadays, as long as you can fit the adapter into the appropriate space.

I like small machines because, well... because they're small. Quadra 605s, 
for instance, can be run at 40 MHz and can have up to 132 megs of memory. 
But they don't have space for either an SCA adapter or an Acard SCSI-IDE 
adapter. I can buy a 50 gig SCA drive for less than a 4 gig 50 pin drive!

>> I saw lockups about once or twice a month on a system which was pushing 
>> around 10 gigs a day. Since I've removed awacs, it has been up 100 days and 
>> done about 4 terabytes of traffic.
>
> OK, so even if the problem was present on my system, it wouldn't've shown up 
> over the short duration of my pounding.  Hmmm.  Anyone compared our awacs 
> with OpenBSD's, maybe they've got it licked?

To be honest, most of the people I know who are running OpenBSD are 
developers; development machines hardly ever do what production machines 
do, so I doubt they'd see the same problems.

It's also very hard to push these systems to their breaking points on 
purpose. Many times I've seen problems with production machines that I 
could not duplicate in a test environment, which is a pain.

>> It'd be nice to fix, but unfortunately I was never local to the machine 
>> when it locked. It's 3000 miles away, which makes it a pain.
>
> Yup.  That does make it more difficult.

That, too.

>> I saw the same problem with a G4 running 2.0, but I put in the motherboard 
>> serial port at the same time I removed awacs from the kernel config.
>
> So, is serial on macppc totally happy and stable?  Last I heard (a while 
> ago), heavy serial usage lead to interrupt issues.

On old world Macs (anything with built-in serial ports, but not the 
Xserves, obviously), using a serial console is tricky - if you type while 
it's sending stuff, the serial console will probably lock. If you type 
carefully while nothing is coming at you, it's fine. There's no way to 
reset the serial port (that I know about) when it locks up except to 
reboot, but the rest of the system is fine.

On the new world Macs, the serial ports default to 57600, and I have never 
seen the serial port lock up. So I think they are completely OK. I can 
even get into the debugger with them! Not sure if that's a new world 
versus old world issue or a 1.6 versus 2.0 issue, though.

John