Re: Editorial review of 7/1/04 version of our draft...

[Nick Maclaren <nmm1@xxxxxxxxxxxxx>]

I haven't seen IBM's patent on this topic, but I have deep suspicions
about its validity.  Let's ignore the fact that such patents are
currently void in most of the world, possibly even in the USA (as far
as I know, no such case has reached the Supreme Court, and there are
constitutional issues involved as well as legal ones).

I sincerely hope that IBM is not patenting ancient techniques with the
intent of demanding tribute from small companies in the future, but
this smells horribly like it.  The exact details of the composite
field are probably new, but that is all that I can see that is.

The main technique (packing 3 digits into 10 bits) is many decades old.
I have been using it since  the 1960s, and have reason to believe that
I was not in the first hundred inventors of the technique!  Using
values of 0-9 (or 0-99 or 0-999) for decimals and 1000+ for out-of-band
indicators is probably newer, but still dates from the 1960s.

I have sent Mike a possibility for a fixed-point decimal format, which
(as he knows) I regard as a much cleaner approach.  Trying to combine
the arithmetic models of fixed- and floating-point leads to a very
messy specification.  For example, ALL of the problems, complexity and
wording associated with cohorts simply vanishes if you split them.  And
binary floating-point is much better for numeric work, anyway.

I am NOT interested in championing it, but am happy to release it for
anyone to use, and I can honestly claim that it is untainted by any
ideas taken from patents (valid or invalid).  But I cannot claim to be
able to prove that, though I think that I could locate enough prior
art to show that it would not infringe the DPD patent (assuming that
were valid).



Regards,
Nick Maclaren,
University of Cambridge Computing Service,
New Museums Site, Pembroke Street, Cambridge CB2 3QH, England.
Email:  nmm1@xxxxxxxxx
Tel.:  +44 1223 334761    Fax:  +44 1223 334679