Can anyone tell me how the method described in this patent differs from standard binary buddy allocation as described by Knuth in 1965? The claims for doubly-linked free lists have obvious prior art (see both Knuth and James L. Peterson, Theodore A. Norman. Buddy Systems. Communications of the ACM, Volume 20 Number 6, June 1977.) The "headers aligned with block start addresses" is likewise anticipated by those two citations as well as every implementation of binary allocation known to me (among those that store headers in blocks). The treatment of non-power-of-two block sizes is likewise anticipated by Knuth, Peterson and many others.
Any comments welcome.