Re: Fw: Meet the $499 Mac.

From: Cameron L. Spitzer <cls_at_truffula_dot_sj_dot_ca_dot_us>
Date: Fri Aug 26 2005 - 19:41:34 CDT

>Date: Fri, 26 Aug 2005 11:55:26 -0600 (GMT-06:00)
>From: charlie strauss <>
>To: Open Voting Consortium discussion list <>
>Subject: Re: [OVC-discuss] Fw: Meet the $499 Mac.

>I agree with alan on the computers for schools argument.

>Mac's make a lot of sense for this. The trend these days is for schools,
>govt labs, and bussinesses to go with fleet agreements with vendors.
>Run an all dell or mac shop. The cost of system administration can far
>out weigh getting a super deal on cousin vinny's home built mini ITX.
>standardization saves more and prevents unexpected incompatibilities.
>Oddly, It's much better for planning to simply know a computer wont do
>something with certainty than to have a mixture of them some of which
>run a mission critical software and some that you discover down the
>line don't. Fleet agreements are a way of taking advantage of this to
>get deeper discounts.

I hope we don't base any public arguments for anything
on a mistaken belief that fleet deals have much of anything
to do with ensuring hardware standardization.
Fleet deals are about locking out the competition.

A few minutes after IBM started selling schematics and ROM
listings (of the PC) in retail stores, a certain kind of
standardization was lost forever and a new kind was established.
PC hardware isn't standardized. An abstract
fuzzy reflection of how the hardware sorta kinda looks through
a few layers of drivers and BIOS overlays is what's standardized.
Connectors are standardized. The electrical interfaces in them
are sort of standardized in an ever changing more-or-less backward
compatible way. COTS hardware is an illusion.

The modern manifestation of this is name brand white box hardware.
What *is* a Netgear FA310? Well, of course, it's a name brand
COTS Ethernet card, right? Sure, as long as you're looking at
it through some API, or looking at the photo on the retail box.
But if you're writing the driver, there's no such thing as
a Netgear FA310. Or, rather, there are several quite different
things. One month they ship it with a Tulip chip.
The next month it might have a New LANCE or 8139
or something. I'll bet the CD they ship with it installs Windoze
drivers for all three, and it loads the right one at boot time.
(If it picked it at install time you'd be screwed when you
replaced a failed card. 3Com has a patent on a driver that
links in the modules it needs for the hardware it actually
finds, automatically and invisibly, at driver load time.)

Video cards and modems are the same way.
Buy the same model Dell twice, six months apart, and you can be
sure you will *not* get the same hardware both times. All Dell
is willing to guarantee is they'll have the same connectors
and run the same applications, provided you use Dell's OS and
driver load. The form factors of high-failure parts like power
supplies and CD drives won't change and the cases will look the
same, but that's all. And it doesn't matter if
you buy them one at a time at Circuit City or a pallet a
week through The Channel under a fleet deal.

That's the main thing that's kept VME and Compact PCI
industrial computers alive. Markets that *really* need identical
replacement parts three years later cannot take advantage of
lower cost PC-compatible motherboards, video cards et al.
They'd have to do a lifetime buy of every assembly, the first
production run, because there is no guarantee Intel or VIA
or Netgear or Asustek or Taiwan Garage White Box Inc will ship you
the same assembly two months later, even if you order the
same part number. Intel's a little more up-front about
their microprocessors. They just go out of production before you're
done designing them in. (That's an oversimplification. Every so
often they notice one had a high volume industrial design win
and they keep making it long after it's technologically "obsolete".
They still make the PQFP 80186. It was really popular for
industrial controls.) That's why you have to design for
the *socket* not the CPU. You're shooting at a fast moving target.
I had to use a PPC instead of a Celeron for my board at Procket,
because Celeron types just come and go too damn fast.
I'll bet it's why cisco uses QED's RISC or whatever.

If our presentation or credibility rests on "common off-the-shelf"
hardware being in some way repeatible, we're going to have to
find a way to deal with this issue. COTS ain't the same way twice.


OVC discuss mailing lists
Send requests to subscribe or unsubscribe to
= The content of this message, with the exception of any external
= quotations under fair use, are released to the Public Domain
Received on Wed Aug 31 23:17:32 2005

This archive was generated by hypermail 2.1.8 : Thu Sep 15 2005 - 11:44:12 CDT