I guess, if I search for "divx", and mark "Gone in 60 seconds divx.avi" for download, Xolox will continue searching new hosts matching "divx" with the same file name and length returned in the Gnutella QueryHit protocol message.

But it looks to me that a less defining query like "divx" is not so likely to include all the same hits as a search for a more refined query with the whole file name. Like when too large hit sets are truncated at clients or just lost due errors or whatever Gnutella network poblems.

So could we benefit of "Searching for new hosts" with the actual file names? Would this be dangerous, because Gnutella does not standardize, how clients do string matching, and so does not madate that "x%&yz" matches all the hits that "x%&" or "&yz" matches?

Or is Xolox already doing this? Looks like it isn't!

ALthough queryhits are not eating up the valuable bandwidth, it could be nice to place more accurate re-searches.

We can (for example) extract the most used names in all downloadhosts from a job and with that refine the search query.

Problems here are:
When you download a whole album of madonna -10 songs-(search madonna),
Then the ideal situation is place 1 requery for madonna in stead of 10 more specific querys.

We are thinking of building a specific query algorithm for jobs that have 'lonely' search-terms, so we will don't have to resend the query with a different specification.

Greets,
Pasman