The congestion-notification conflict

Posted Mar 27, 2019 18:11 UTC (Wed) by ajb (subscriber, #9694)
Parent article: The congestion-notification conflict

I took a look at the patent in question (at least the US version) and noticed the following, which gives some hope that the patent may not be as much of a blocker as this article suggests:

All the non-dependent claims in the patent (that is, claims 1, 14 and 22) seem to assume use of the 'proportional-integral controller', (a object from control theory which is used in AQMs). But, the dualQ L4s draft (draft-briscoe-tsvwg-aqm-dualq-coupled) gives, in Appendix B, an alternative to this called 'Curvy RED'.

( All the claims of a patent are either non-dependent or depend ultimately on one of the non-dependent claims, so when all the non-dependent claims rely on something then it is essential to the patent as a whole).

In email on the TSVWG list, Bob Briscoe, one of the developers of L4S, clarifies here that the dualQ part of L4S, to which this patent applies, is intended to be a framework into which any AQM could be dropped.

Accordingly, it seems to me much less likely that it would first appear, that Linux could be locked out of high-perf networking if L4S is the winner in this debate.

Note that I am not a lawyer, and it would be good if others could read the patent and post if they concur or not with my reading of it.

It is still less than ideal that the DUALPI source, which is posted for evaluation of L4S and for adoption into linux, has this patent hanging over it. The developers might be well advised to post a version which avoids the PI controller - unless they believe that ALu can be persuaded to licence the patent freely to OSS.