[conspire] Re: return to a state of grace

[Rick Moen]

Daniel Gimpelevich <daniel at gimpelevich.san-francisco.ca.us> wrote:
> What makes you think that'll ever happen? I looked up the specs on that
> SMC card you found:
> http://www.smc.com/drivers_downloads/library/8216-DS.pdf
> Note the Mean Time Between Failures. I haven't looked up the 3Com cards,
> but I don't imagine they'd be any less reliable.

I don't know what specifically kills old ISA cards, but they do
occasionally shuffle off this mortal coil.  Could be power spikes.  Or
heat.  Or both.

For all I know, that pair of old 3C509B cards may outlast everything
else I have:  My point is that the potential hassle and lack of
long-term reusability of even a really good ISA card means deploying
them going forward even as replacements just isn't worth the trouble.
(I knew the days of ISA were officially over when Adaptec's price for
their flagship AHA-2940U PCI SCSI card dropped below that for the
AHA-1542CP ISA card.  And of course, for the last couple of years, new
motherboards don't even come with ISA slots.  Soon, I expect that many 
new ethernet switches will lose the ability to negotiate 10 Mbps.)

But part of what I valued about the 3Com cards, in their day, is that
they satisfied Moen's Law of Hardware beautifully:  When in doubt, use
what the programmers use.  In other words, if you use hardware that
coders favour, then you'll tend to enjoy superior driver quality and
performance.  And drivers for the 3C509 series were outstanding on every
OS, for that reason.

> BTW, that SMC card's configuration has an option to enable or disable
> "Zero Wait State Operation." What exactly is that, and which is the better
> setting?

Means it will expect to try to pump data in or out on every clock cycle,
instead of deliberately being idle every other cycle to allow slow bus
circuitry to catch up.  Zero wait state = On is faster, assuming
reasonable host (motherboard) circuitry.