Firstly that the existence of both a free and a for-fee copy of the same article seems a bit unlikely. I can imagine a case where this could happen. (Private pre-publication on the author’s personal web site before publication in a journal.) Not exactly the same thing bibliographically, but the content is the same, so ’same enough’. But, in general, the only way I can see an originally for-fee article becoming free is if it moves into the gray area of fair use copying and posting. And for that we can hold our collective breath and see what Georgia has to say to the Publishers.

The second, more technical, interest is how this free copy “happens to be found”. I agree with Jonathan that finding such animals is tricky. They obviously don’t exist in the search indexes of the major content providers (sort of bad for business to offer the same thing for money and for nothing - which would _you_ choose?). That means to find them you have to look in lots of small places - personal booklists (article lists?), project repositories, personal websites, and so on. These small locations are initially unknown and then difficult to connect to (quite possibly home grown systems), and then only likely to be of use for vertically specific searches (they are very ‘long tail’). So you end up federating across the generic web search engines to find search engines which should have the sort of content the user is searching for. Hurray! we have meta-meta-search. So far this form of source discovery is very arcane and seems to be a bastion of those totally unreliable finding aids called humans. Until the semantic web abounds, I think it will stay the province of “I know a site that has …”

A big barrier I’ve come into, as I try to write this software, is the difficulty/impossibility of finding/identifying a free copy of a paper that someone found in a search, that may exist online in both free and for-pay versions. There’s really no way I can find for my software to discover the free version and know that it’s a free version. So the “happens to be found” part is trickier than it sounds.