[If you're looking for anything serious, don't waste your time with my musings; instead look at MacKie-Mason's papers at U Michigan who has spent much more time with much more aspects of accounting on the Internet than I have. Or look at the material on Market-Oriented Programming, also at U Michigan.]
These things are entered into a searchable database.
All participants can search and browse the descriptions of all available resources. They can see the description and the access state - how many accesses are happening per time, how many are left; what they don't see is the URL.
If they like the descriptions, they buy accesses to the actual documents with play money, tokens. Half of the tokens is "paid" to the person who entered the URL; half of it is deducted by the server maintainers to keep the system hungry.
URLs are sold only as many times as they can sustain accesses. If the URLs don't work, the buyers can complain back to the system and get their money back (if necessary, from the initial provider's account); the URL is removed from circulation.
If you run out of tokens, you can generate new ones by mirroring existing resources that are in high demand; you buy a one-time peek at something and use that peek to copy it, then sell accesses to the copy.
Since the accesses are counted, you can safely remove the copy after it has been accessed by the people you sold it to.