The program would wonder the same things as you: what are the filenames and where are they located. There is no way of simply looking up such information, so it would try a.mp3, aa.mp3, aaa.mps, b.mp3, et cetera indefinitely, eventually reaching www.ted.com/zzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzz.mp3 before moving on to www.ted.com/a/a.mp3 and repeating the process. In short, it is a search that could actually exceed your own lifetime.

However, if all of the MP3 are directly linked to somewhere on the site, and all of the pages are somehow traceable back to a single page, and countermeasures have not been enabled to deter/block web spiders, it would be possible to search every page on the site. Given the sudden bandwidth consumption, however, the operators would not be appreciative.

My suggestion: Just contact them and ask for a bulk download.

John