* makes class='internal' default, saving lot of bytes
* changes all <em> to <i>
* changes all <strong> to <b>
(nobody showed actual case where it breaks something, so ...)
* a few one-byte optimalizations here and there
Improvement is not going to be big - using relative links would have
bigger effect that all these combined - but it's trivial and shouldn't
break anything.
It definitely needs testing (but I couldn't apply it to instalation at
test.wikipedia.org, as it's a bit far from CVS now), especially:
* how it changes rendering with underlining turned off.
css code is a bit suspicions here, but results look
fine to me.
* does changing <strong>/<em> to <b>/<i> affect anything ?
it shouldn't, and if it does we need to fix css code, not
keep wasting bytes version.