Apparently, he didn't directly open source it. He approved its inclusion in the Open Refine package. It's pretty obvious that he didn't understand the consequences of releasing his code in a project licensed by BSD. I can understand his frustration, probably because he was told that his conditions were covered under the BSD license. Given that he sells the same code as a commercial product, this was obviously not his intention.

However, he failed to grasp some basic IP concepts, especially wrt US law. Algorithms are not copyrightable, and most are not patentable (especially after Alice v CLS) under US law. However, there is solid precedent for similar systems being eligible for patent protection (soundex, other patents that cite his systems as source material, etc). It does not appear that he applied for this protection at this time.

Moreover, under the modification clause of the 3 clause BSD license he used in his source file, there is a significant amount of commentary suggesting such a rewrite would be permissible under the terms of that license. However, re-licensing it under MIT would on the surface not be permitted.

EDIT: he may have applied for a patent. This could preclude the project from including his system. However, I'm not a patent lawyer, so I don't know how well this would hold up in international implementations (github being a US entity makes this point relatively moot).

http://www.google.com/patents/US20090043584