I do it with  tagsoup and XSLT. Why would you want to do it with anything
else if you know XSLT (of course if you don't .....).

I am aware of one argument against where the scrape needs to be highly
performant (perhaps because it is done in real time).

I am keen to hear any others.