Feedback on the RAND policy

To whom it may concern,

Section 4(a).2 "portions of an implementation that are not required by the
Recommendation."

If I understand this correctly, this is giving extra weight to the usual
indications of "MUST", "SHOULD", and "MAY", and of course, their negatives.
If something "MAY" be implemented, does that make it required by the
specification?

Either the policy must be expanded to explicitly require in the
Recommendations declarations of which of "MUST", "SHOULD", and "MAY" are
"required" by the definition under this clause, or this clause should
clarify the use of "MUST", "SHOULD", and "MAY" with regards to "required by
the Recommendation".

My personal preference would be that anything specified in a recommendation,
whether as "MUST", "SHOULD", or "MAY" should be included in the arena of
essential claims.

Section 4(e).1 "shall be available to all implementers worldwide".  Indeed,
I think this is too narrow, in that "implementers" is undefined.  I am not a
lawyer, but what is wrong with "shall be available worldwide to all".

Since part of the point of the W3C is to promote standards that will be
adopted, I am deeply troubled by the underlying premise that RAND licenses
might be acceptable.  Specifically, the definitions of "reasonable", and
"non-discriminatory" are problematic.  When sliding up the scale, is $50000
a year in licensing fees reasonable? How about $100,000, or even $1,000,000?
The smaller the company involved in the licensing, the less "reasonable"
even a $50,000 fee seems.  Or a license could be written on a "per-copy"
basis, at $5 a copy - which could be unreasonable for a product (like email)
that might otherwise go for $30 - except for the "reasonable" fee added on
to it.  Does non-discriminatory mean that the licensing fees are lower in
countries with smaller GNPs?  It probably should.

Where does this leave programs that rely on "sponsorship" such as an a
Eudora or Opera, wherein they make their revenue from advertisements in
their products?  Even a minimal licensing fee for these products would
prevent the implementation of a feature covered by a so-called "RAND"
license.  And maybe that is OK, but it still seems indicative of the problem
to me.

Another, somewhat unrelated thought that I have - what about a sunset
provision as at least an option within the RAND license?  The problem with
patents, among other issues, is that they last for so damn long by software
development standards.  The software I write today is completely different
than the software that I wrote back when I started fifteen years ago - and
yet a patent in effect back then could still be in effect now - and still
slowing down the pace of development if it applied to what I was working on.
What if there was a sliding scale, so that "early adopters" of a standard
would pay licensing fees, and the rest of the world could wait a few years
to get it under an RF license?  I could imagine a 1, 2, 3, 5, and 10 year
"sunset" on the RAND licensing provisions (from the date that the
recommendation goes public).

I am troubled by the implications of this, and I hope that the W3C takes
care to heed the advice of others, smarter than I, who might have valuable
input on the matter.

-Eric Johnson
TIBCO Software

Received on Thursday, 4 October 2001 12:06:36 UTC