A facetiously named workshop on REST was held at WWW 2010 yesterday. After a day of interesting papers on the distinctions of REST and non-REST architectural styles for Web Application Programming Interfaces, the thing which struck me as most important was the recognition that (a) although it may affect them deeply, users don't care about the finer points of the architecture so long as they get an API that works and (b) there is something wrong in what the community understands about the theory and application of links within the payloads of Web APIs.
I did not mean to be impertinent, but I had to remind the WS-REST panel at the end of the workshop that much work had been performed in the SGML world on the subject of addressing and linking, in particular HyTime. The panel members were quick to acknowledge the prior work. One subtle rejoinder made was that HyTime was developed for interactive applications, not for programming interfaces.
I let that comment slide when I should not have. While ISO10744:1992 is hundreds of pages and not necessarily the standard to solve their immediate problems, HyTime was developed to provide guidance on markup language architectures, not any specific applications. A substantial fraction of those pages are relevant to the linking and addressing problems however, and much research effort was put into thinking about the stability-over-time-and-location problem faced by REST adherents today. Standards that go around, come around.