The open patent idea is actually quite old and has traditionally been practiced by consortia of research-oriented companies, and increasingly by standards bodies. These also commonly use open trademark methods to ensure some compliance with a suite of compatibility tests, e.g. Java, X/Open both of which forbid use of the mark by the non-compliant.
Thus the model already has a strong legal framework. Patent improvement licensing is already practiced by some global institutions, notably the government of China and MIT. Each of these manage a large patent portfolio and often require improvements to be licensed back as part of the original portfolio, although this is not the default license.
Critics question whether the promoters of truly 'open' and mandatory improvement licensing, having spent most of their lives opposed to software patents, can actually attract donors of patents, or would actually participate in a process that they claim to despise. This criticism probably focuses unduly on the personality and ideology of Richard Stallman, who has nonetheless sought to solicit donors for such schemes. He found, not surprisingly, that software patent holders were not so interested in talking to him.
However, despite confrontation between Stallman/GNU and the patent system, the open patent movement got going and attracted some support. It remains to be seen if it can become a major phenomenon - patents are difficult and costly to obtain and require extensive documentation, unlike copyrights. Patent rights on software and on life forms are controversial and many activists believe that they can successfully prevent such patent rights from existing at all, and so would be less inclined to patent and contribute to any such portfolio.
See also: copyleft, open content