🚀 go-pugleaf

In article <C0JoJ6.LG6@zoo.toronto.edu>, henry@zoo.toronto.edu (Henry Spencer) writes:
> In article <DERAADT.93Jan7120536@newt.newt.cuc.ab.ca> deraadt@newt.cuc.ab.ca (Theo de Raadt) writes:
> >Why the heck did Sun not put better serial chips in their new machines??
> >Isn't it obvious by now that the Zilog chips SUCK?
> 
> Probably because Sun no longer has any competent device-driver writers...

How competent can a device-driver writer be if their job is to stick themselves
between a rock (interrupt-per-character serial chip) and a hard place (long
interrupt latency, expensive context switches, complex environment)?

The real problem is that most serial chips are still late 1970's creatures,
designed to work with 8-bit microprocessors in bottom-dollar applications.

If the bloody things had some decent buffer ring management in them (Cirrus
Logic has some multi-serial chips which do), then they could be made to work
nearly as well as Ethernet controllers do.  Except of course for the fact
that most serial interfaced devices are cheap and un-designed junk, so people
expect feats of miraculous intuition in flow control to keep them from
overrunning each other.

Ethernet: the only decent low-speed serial interface.  Give a design you love
an Ethernet port today.

Brad Yearwood   brad@optilink.com    {uunet, pyramid}!optilink!brad
Petaluma, CA