[conspire] Ah, it's an Atheros-based PCI card

Rick Moen rick at linuxmafia.com
Thu Sep 8 11:56:16 PDT 2005


Quoting Ryan Russell (ryan at thievco.com):

> Michael T. Halligan wrote:
> > In my experience, that is only 1/2 true. They'll change their
> > chipset entirely (a full vendor switch), and increment in a funny
> > way.. Like I remember there was an FA310 .. then an FA310A .. then
> > an FA310b then an FA311 .. all of them 10/100 ethernet cards, all of
> > them completly different chipsets.. all but one or two of them used
> > the tulip/decchip driver, but the tulip driver that worked 3 months
> > ago on the old chipset wouldn't work on the new one, and you'd have
> > to update.. A real pain. Must be "Just in Time"
> 
> All good examples of what I was talking about.  When I ran into it,
> there were no identifying revision letters on the box.  The A/B only
> came into play when you got to the Windows drivers.  The Windows
> drivers on the included CD would always work with the included card,
> and any earlier versions.  The problem came up when you tried to buy
> one in a store or via mail order.

Succinct description.  I remember that debacle, now.  That was the
incident that gave us the classic Donald Becker line about the related
conspicuous display of ineptitude from Bay Networks / Netgear:  "Hey,
all of these media selection problems are trivial if you think like an
MS-DOS programmer."

See:  "Netgear 310" on http://linuxmafia.com/kb/Hardware/

As someone who's worked both the hardware and software sides of the
fence, I'm conflicted.  On the one hand, as users and driver folk, we'd
like hardware vendors to clearly warn/disclose all significant chipset
changes.  On the other hand, life as a hardware VAR sucks:  Suppliers
will make _their_ own changes without notice, or will simply discontinue 
key parts assemblies/components you were relying on for your three-year
roadmap to your customers.

The unfortunate fact is that there's a continuum -- a spectrum of
variability -- ranging from outright breakage caused by substitution of
a radically different chipset (e.g., when Adaptec's 1542 ISA SCSI card
series went from CISC to RISC) to subtle variability over time that
sometimes matters and sometimes doesn't (e.g., Diamond Multimedia's 
habit in the late 1980s of using the RAMDAC du jour in its accelerated
video cards).  Knowing that "the DOS driver still works" didn't signify,
because, as you describe, that DOS driver was in those cases _not_ the
same, having been up-revved to compensate.

At $PRIOR_FIRM, a well-known Linux hardware vendor, we certified
Mylex-based hardware RAID arrays with certain hard drive models -- which
promptly disappeared.  Keeping those arrays reliable was a constant
struggle with upstream vendors, where we were forever having to run
everything through test matrixes _yet again_ because somebody had made
an unannounced parts substitution -- and where only certain BIOS
revisions (and definitely not the latest of those!) were certified to
work.

Looking from the perspective of the hapless Linux user, all you can
really do is bear in mind that driver functionality is subject to
breakage from diverse causes, and keep an eye out.

-- 
Cheers,                 Katrina's Law:  Any sufficiently advanced incompetence
Rick Moen               is indistinguisable from malice.  
rick at linuxmafia.com                           (coinage attrib. to Paul Ciszek)




More information about the conspire mailing list