🚀 go-pugleaf

Regarding the need for a separate serial-port processor,
louie@sayshell.umd.edu (Louis A. Mamakos) writes:

> This isn't always the case.  If you choose some serial encapsulation
> that the 8530 supports (like HDLC), you can arrange to have it
> interrupt you and terminate the DMA transfer when it receives an
> entire frame.  This should work *better* than the normal UNIXy sort of
> transfers.

No, it isn't *always* the case, just *usually*. If you have the luxury of
designing the machines you want to talk to, and you can afford to put
HDLC USARTs in them, you can do the HDLC framing trick. But let's be real:
we usually don't get to design the machines we talk to. They usually have
lowest-common-denominator UARTs in them, and speak higher-level protocols
designed by a cretin with sadistic tendencies. I want my serial-port CPU,
and it had better not be some overpriced 68000 derivative with an interrupt
response and context switch measured in geologic time. I've spent several
years of my life getting paid to dance around serial ports with these
monsters, and my toes are real sore. If there are any silicon jocks out
there, what the world needs is a faster 8-bit microcontroller with a sensible
architecture, not another 32,48,64-bit showboat to further dazzle and confuse
our managers.

Louis and I agree on one thing: NeXT needs to release docs on writing serial
drivers. I shrieked when I found out that NeXT doesn't want us in there: the
drivers they give us just won't do.

Steve abell@netcom.com