[ NNSquad ] Re: David P. Reed, nnsquad member, kicks ass at the FCC hearing!!

[Barry Gold <bgold@matrix-consultants.com>]

Jeffrey S. Young wrote:
> WRT standards, I come at this from a provider's point of view:
>
> IMHO, the standards making process in the Internet is a bit different
> than that of other technology standards bodies (perhaps compare
> the IETF and IEEE).  The IEEE publishes real standards and if your
> equipment doesn't abide them, it doesn't work with similar equipment.

You obviously never had to write a device driver.

I maintained the ATA (IDE) driver for Solaris for two years.  The 
standard is pretty straightforward.  Load up a few registers and let 'er 
rip.  You need to pay attention to which modes are supported (as 
reported by the response to IDENTIFY DEVICE).

What makes the driver complex is that about half the devices on the 
market _don't comply_ with the standard.  Some controllers use a common 
control register for both devices, so you can't overlap operations even 
though the standard claims you can.  Some devices claim they do DMA, but 
it doesn't work -- or it doesn't work at the speed they claim it does. 
Almost the entire driver is special cases, recognizing specific devices 
and working around their glitches.

> Internet 'standards' are more or less recommendations that are either
> deployed or are ignored.   The two communities that influence what
> becomes a true IETF standard and what is relegated to "informational"
> status are the ISP's and the user community.
>
> In the ISP community, if the standards solve a problem, give an
> ISP a competitive advantage, they are deployed, if not they are ignored
>  (i.e. BGP4 vs. sBGP or soBGP, DNS and DNSsec,  IPv6 and so on).

The same is true of the IEEE standards.  If there is a higher adherence 
to them, it's because companies that don't comply find their product 
doesn't sell.  And if they can gain a competitive advantage by doing 
something else, they probably will.

Those of use with long memories may recall a controversy when it was 
discovered that one company (Sun Micro?) had set their backoff times to 
be 10% shorter than the standard, so that their devices were more likely 
to win "collision wars" and hence showed better results in performance 
benchmarks.

The bad publicity _did_ get them to return to standards compliance, so 
overall I guess the 'standard' worked.

> Deployment wins!

Equally true everywhere.  I'm not a hardware engineer, but I talked to 
several.  All agreed that Sony's Beta system was technically superior to 
JVC's VHS.   Ten years later it was almost impossible to buy a Beta 
machine -- VHS had won the deployment battle.

I don't know if this was just because VHS allowed you to put more on a 
tape, or because of some unfortunate design decisions in the early 
Betamax recorders (e.g., connecting the capstan to the drive motors with 
a rubber band, which would gradually stretch and then start to slip, 
producing unreadable tapes).