There are a couple of things that could be improved in the extensible sequence protocol, which is now documented (thanks to Jan Moringen):

• a sequence-element-type function would be useful for sequences which wrap vectors (and for such toy examples as the DNA-as-two-bits sequence);
• as currently implemented and documented, the base class from which user sequence classes should inherit is bare sequence. This is wrong from a compatibility point of view (what if a competing extensible sequence protocol comes along)?