Sunfish patents (was: Vienna proposal...)
It seems to me that Mr. Subby's characterization of Prof. Arnold
Neumaier's presentation is grossly unfair. As far as I can tell
Arnold went out of his way to try to phrase things in a way that
does NOT infringe on anybody's intellectual property. Was it
really necessary to accuse him of reverse-engineering a protected
method?
What the Interval Arithmetic community was hoping to get from the
principals involved was clarification on how best to proceed in
this manner. For example, is it the identification of the simple
linear interpolation as a primitive that is at issue? Or is it
the "obvious" implementation thereof in a manner that obtains tight
bounds? Saying that "anything that produces the same result" is
protected doesn't make much sense, as surely the invention is to
have found a clever and significantly more efficient way to find
an optimal enclosure than brute-force computation of all possible
combinations of inputs...
The proposed standard makes the interpolation primitives optional
precisely so as not to tie the standard to required patents -- if
that turns out to be an issue. But now it seems that anything that
performs interval arithmetic is at risk -- or perhaps even anything
that performs arithmetic at all?
I should also note that interval polynomial evaluation as such is not
addressed directly at all in the Vienna Proposal. The techniques that
can be used to do that efficiently are mentioned only by reference, and
those references include the Sunfish patents. So where is the beef?
Michel.
(NOT speaking for IBM of course, though being at IBM does make me aware
of the need to protect Intellectual Property.)
Sent: 2008-12-22 22:43:43 UTC