I think the idea would be worth a trial and see if it is useful. As others
have mentioned, inline source code and URLs could cause problems. I'm also
a little concerned (until proven otherwise) about the possible processor
usage. OTOH, people were concerned about the processor usage for OCR, and
that turned out to be largely a non-issue.
I don't personally have any real feel for whether this is a good idea or
not. Generally that means to me that it should be tried (maybe several
different ways) and see if something comes of it.
A variant of the concept that seems that it would hold promise would be some
form of "obfuscation index". This might be based on the number of
mis-spellings, or the amount of punctuation or density of punctuation or
some such. But to get the numbers right, you might want to know the base
language as a starting point.
Loren
