Joe Seigh
8/11/2004 1:16:00 PM
David Holmes wrote:
>
> "SenderX" <xxx@xxx.com> wrote in message
> news:gHcSc.263867$JR4.242375@attbi_s54...
> > > Basically, using compare and swap to push onto a linked list. Plus a
> way
> > to do
> > > versioning to avoid the ABA problem it looks like.
> >
> > They have really discovered something radical and innovative here...
> >
> > I bet their R&D for lock-free algorithms == Reading this group!
>
> How the hell do we stop this insanity? :( Is anyone at the patent office
> actually technically competent?
>
There's also this patent, 6,178,473, "System for selectively incrementing a count number of
an associated node only when the node is put in use in conjunction with a successful
compare and swap operation", which I've mentioned before. The only difference between
this and IBM's version is they increment the version count on pop whereas IBM's version
increments the version count on push.
Part of the problem is that there is no good way to search the patent database for related
art. I suspect there are commercial databases set up for this but no public ones. Plus
Linus is going around recommending that no one read patents because of liability concerns
so public knowledge in this area is going to be limited only to people who've had direct
knowledge of the technology in question.
I know about one patent accidentially and one because it references one of my patents. I've
looked at relatively few patents so based on that sample size, there's probably dozens
of patents at a minimum on this technique alone. We just don't know about them.
Joe Seigh