Stripping out non-*ML tags?

I have an application that's using XSL to render text that was
originally put into an online message board system.

Unfortunately, the message board system uses its own proprietary "forum
code", which people who've used such systems will probably be familiar
with.  IE [b] for start bold, [/b] for end bold, etc.

Strictly for readability, I need to strip or at least minimize the
amount of that stuff that gets past my XSL.  I would be perfectly happy
to replace, in regex terms, \[[^\]]\] with "".  That is, just delete all
of the tags, so that "[b]text[/b]" becomes "text"

Is there any reasonable way to do this in XSL?  By reasonable, I mean
something that won't clobber performance.  If necessary, I could move
the stripping to the DB side, but I'd really rather not do that.

Thanks,
Brooks Talley
FRNK Technology Group