Hi All,
I'm having some trouble in the query department - i don't understand why querying by URN is bad in Gnutella, and what replaces it.
Say i want to find more souces for the file i'm trying to download - the simple way for me is send a query request with its URN (sha1) - i saw that LimeWire does not support this (only 1 hop) - why send a query by keywords if i'm getting alot of irrelevant answers? i know other P2P networks use the search by a unique key to limit results for mathces only.
Someone (maybe kun) told me here in this forum that querying by URN pollutes the network but i don't understand it, and that Gnutella will use DHT - so, is it used? If so, does it hash keywords as well or there it uses URN?

What's wrong with querying by URN and is it possible to do that in Gnutella?


Thanks for any help,
Snayit

use query by urn - they wrote it in their code - it can go one hop but it doesn't continue further. So i guess it's impossible.

Is there anything that replaces query by urn and allows searching without keywords but given a urn - find locations for it? Maybe i can use Magnet links for that purpose? DHT maybe?

LW will drop queries by URN. Expect it to be back only when a DHT is up and running. Many servents were abusing from requeries by URN and were overloading UPs = search results sucked.

Ciao

__________________
Liens d'intérêt /Links of interest:
Gnutellaforums en français /The House's rules you have to respect / First search the forum, then create a thread / Free software alternatives! - Logiciels alternatifs gratuits!/

queries if it contained only the urn. When i connect to BearShare for example i receive alot of queries that contain only urn:sha1. But LW servers don't pass my queries and i know that they are a large portion of the network. Are they about to integrate DHT?

What are my means of searching if i want more sources and have only the urn? can i search by urn not using the query? can i retrieve the file name in any way?

If not it means that i should only rely on the download mesh in order to get more sources.

You have to rely on dl mesh but you can also help the user make a requery based on key words of the file being downloaded (when there is no source known)

Bearshare behaves differently. LW is integrating azureus DHT so...

Ciao

__________________
Liens d'intérêt /Links of interest:
Gnutellaforums en français /The House's rules you have to respect / First search the forum, then create a thread / Free software alternatives! - Logiciels alternatifs gratuits!/

If i need to retrieve the keywords for the file, knowing only the urn - how can i search for the file name or a part of it in order to have keywords (if it's possible in the stage of DL)?

Not sure the DHT will help me here if they still forbid querying if i only have the SHA1 - they can implement DHT that uses keywords only.

LimeWire does not want you to query the Gnutella network for alternate sources in its current shape. Ignore the "search for keywords instead" red herring. First of all, if you only have the SHA-1 that's strictly impossible, thus nonsense. LimeWire only "encourages" searching by keywords because this makes it much more likely that the query terminates soon.

Wasn't this sufficiently explained in this thread:
http://groups.yahoo.com/group/the_gdf/message/22146

If not, then read this older thread:
http://groups.yahoo.com/group/the_gdf/message/20774

The sole purpose of adding a DHT to Gnutella is making searching by SHA-1 (or possibly any other unique identifier).
efficient and thus support it again. I don't think anyone intends to use a DHT for keyword-bases searches.

I'm not a troll, please don't be angry with me - just trying to figure out lots of new-for-me stuff.

If, for the sake of the argument, i DO have only the urn (got it from BearShare in a query - sent me only the urn-SHA1 with "/") and i want to download this file myself - i won't be able to do so cause i don't have its name and won't have sources.

My first option for getting sources will be to address the user that asked me about the urn - maybe it has the file and i can use it as a source, but because the queries are coming from UPs he's not really the source. So now i have only the SHA1 and no sources and i can't get the file. So query by urn - no such thing now, but can i get the file name in any way?

Ok. Promise this is my last message on that matter. Sorry if i upset anyone, and thanks for still answering me.

I'm not upset but it's annoying that this thread was running in cycles just like some threads over at the GDF. I assume that's because it's not fully understood what a DHT is or why Gnutella would benefit from one.

Anyway, one weak option is to use a Bitzi lookup:

http://ticket.bitzi.com/rdf/urn:sha1...NAE52SJUQCZO5C

Of course it knows only about a very small subset of all available files, it's a single point of failure and the information it provides is not trustworthy. So it's nice to have but no solution.

Also you cannot expect that any given SHA-1 actually has a corresponding existing file. And even if this file exists, it may not be available from any peer when you're searching for it. If all peers use a common DHT, it is at least easier to determine whether a file matching the SHA-1 is available or not.

Something that would be kind of interesting is running an Ultrapeer and collecting data from the routed traffic. After a while you're probably able to map most SHA-1s of popular and many other files to such information as filename (actually a list of), filesize and so on - depending on what information you record.

and it didn't find it - it has a few files. The thing you metioned - checking the net for popular files is exactly what i want to do (i aim to please) but you gave me another point of view for that (keeping the file names if i have them).

I still haven't succeeded being an UP, so i'll try investigating in that direction maybe. Also, i'll try checking if maybe when i DL a file it gives me its name in the extensions of the protocol.

Thanks alot.

Many clients (BearShare, Gtk-Gnutella, LimeWire) emit a Content-Disposition header when you download a file, so if you have a working source, that would tell you one of the used filenames. You have to be a bit cautious with the encoding for non-ASCII filenames though; LimeWire and Gtk-Gnutella should always give you UTF-8 encoded and NFC normalized filenames. BearShare probably doesn't. They also use a slightly different format, for example, Gtk-Gnutella explicitely declares the filename being UTF-8 encoded, the others don't. You'll see the effect of this when you're trying to download something with Mozilla - well LimeWire and BearShare probably block browsers anyway but you get the idea.