That's why I added my disclaimer "if you're trying to create a semi general purpose solution" (sorry about the typo). Very simple things can be done with a regex. Also if he knows exactly what his data looks like and this is a one-off job a regex is quite likely to suffice. I assert that unless you have some experience to make a good call it's safer to err on the side of using a parser for X?(HT)?ML where a regex might have sufficed, though.

But in this case I have no idea if he is even parsing markup at all or just something that happens to look like it. So instead of giving a possibly ill-advised suggestion I chose to just raise awareness about the issue and leave it at that. Sorry to sound like a broken record, but that's because I'm responing to the broken record of a question that keeps coming up.

Makeshifts last the longest.