Thanks for bringing this up Erik. It really does seem to be preferable
to me to treat these tiles as web resources in their own right, and to
avoid treating them like resources that need to be routed to with
OpenURL. It is also seems preferable to leverage RESTful practices
like using the Accept header.

I wonder if it would improve downstream cache-ability to push parts of
the query string into the path of the URL, for example:

  http://an.example.org/ds/CB_TM_QQ432/4/0/899/1210/657/1106

Which could be documented with a URI template [1]:

  http://an.example.org/ds/{id}/{level}/{rotate}/{y}/{x}/{height}/{width}

I guess I ought to read the paper (and refresh my knowledge of http
caching) to see if portions of the URI would need to be optional, and
what that would mean.

Still, sure is nice to see this sort of open source work going on
around jpeg2000. My nagging complaint about jpeg2000 as a technology
is the somewhat limited options it presents tool wise ... and djatoka
is certainly movement in the right direction.

//Ed

[1] http://bitworking.org/news/URI_Templates