On Mon, 25 Feb 2002, Billy Tanksley wrote:
> From: Daniel Berlin [mailto:email@example.com]
> >4. We now range encode (almost-arithmetic encoding, but not
> Are you sure? I don't see any patents, but arithmetic encoding is VERY
> heavily covered by patents, and this is VERY close to arithmetic.
Yes, i'm aware.
However, reading the arithmetic patents, they all seem to specifically
refer to renormalization in terms of bits.
Range coding described in paper published in 1979 by IBM UK, and they
never patented it. I checked this through patent searches myself, rather
than taking someone else's word for it. As a law student, I have
access to the types of databases necessary to do good patent
searches. Not that any of what i say should be interpreted as legal
advice. I'm not a lawyer yet, my opinion is as legally useful as a cat
hitting randomly on keys generating messages, etc.
I'm also a former IBM research person, so i fired off some email to the
guys in charge of patents just to be on the safe side.
Even, absolute worst case, if we're all wrong, and they had patented it
before 1979 (which would be quite surprising), and it took 5 years to
grant the application, it would have expired at the beginning of the year.
AFAICT, it's not covered by the other arithmetic coding patents (which all
renormalize in units of bits).
Their patent on the basic means of arithmetic encoding, #4,122,440
(granted in '78), is expired now, the rest all specifically seem to cover
renormalization at a bit level.
> I hope we can use this; the results are unimpeachable. BTW, where did you
> get the PD implementation? The only one I can find is GPLed.
Dmitry Subbotin had a PD implementation in PPMd, Mikael Lundqvist
converted it to C and extended it a little bit.
I took his, and removed the stuff we don't need for subversion, and make
it able to work on svn_stringbuf_t's.
To unsubscribe, e-mail: firstname.lastname@example.org
For additional commands, e-mail: email@example.com
Received on Sat Oct 21 14:37:09 2006