Wouldn't the effects of recursion will be lost in the static,
compared to the effects of loading the doc into memory to facilitate
tree processing? Even if you are doing some persistent-ancillary-
info trick to do a virtual tree, in my experience for very large
docs you really have to wrangle memory carefully. It seems
really counter-intuitive that the stack & local variables overhead
caused by recursion is going to get you before one of these
other things. Unless of course you recurse in some huge
sloppy badly-written routine with lots of local junk.
BTW, what languages can be relied on to do tail recursion?
Also, shorter algorithms are better. -Tim