On Wed, Nov 16, 2005 at 11:10:41PM -0800, Chuck Silvers wrote:
> well, for my day job I implemented a version of this myself several years
> ago (and spent forever tweaking it to not cause problems with various
> single-threaded patterns), and some other guys later tossed that and
> wrote a different version.  however, neither of these is open-source and
> to make it worse, both algorithms are patented, so we'll need to do
> something different.  I'll look up the patent numbers so you can see
> what we need to avoid.  I have some ideas on other ways we can do this
> that should avoid the patents, but I'll let you think about it for a bit
> separately first, hopefully you'll have some fresh ideas.

the US patent numbers I mentioned above are:

6070230	Multi-threaded read ahead prediction by pattern recognition
6789171	Computer system implementing a multi-threaded stride prediction read ahead algorithm
6915404	Computer system implementing a multi-threaded stride prediction read ahead algorithm

(I'm not sure why there are two for the second one.)

the full text is available at
http://patft.uspto.gov/netahtml/srchnum.htm

-Chuck