From: "Mike Champion" <>
> Well, not to touch on yet another permathread, but to the best of my
> knowledge as a looker-over-of-shoulders of real implementers, this is not
> "premature" -- it's the hard-won knowledge from a lot of profiling that the
> buffer management needed by the possibility that some token will be an
> entity reference that needs to be expanded is a significant bottleneck in
> many implementations.
Oh, so their optimisation is to assume that references will expand
to the same (or fewer) characters. I guess one could assume that
all element names had 1 character, and then complain about the
inability to use this splendid optimization because XML says we
can have more than one character in an element name.
Cheers
Rick Jelliffe
