Conclusions

Doing a decent job of a ``simple'' matter of proposing standard ways of constructing feature identifiers thus turns out to involve interelated aspects of dynamic schema discovery, metadata granuality, formal version control semantics, distributed/replicated unique naming systems and a lot more besides. Despite the temptation to thow up our hands in horror, there does seem hope that very tightly-defined and narrowly focused feature handles may yet provide some usable functionality which is worth the effort of implementation.