it looks like textual content of attributes and cdata
is handled by the inline strings, but tags have to use LITERAL, which uses an
offset into a table.
Dang! Though I suppose if we wanted to pursue this we could make our own variant that allowed inline tag/attribute names.
Another worry is that it doesn't appear the binary format can
extend to multiple bytes if needed - more than 64 unique tags would seem to
overflow the dictionary.
Isn't that why they have the whole page-swapping thing to support multiple single-byte namespaces for tags?
Internally loaded (shared library) components of the server communicate via
structures which are centered around DOM-like XML nodes
OK, that sounds reasonable. So I don't see the issue -- the binary protocol could easily be parsed into exactly the same DOM data structures as the textual format. All it requires is a parser with a different front-end, which is what I meant in my last post.
�Jens
