Gnutella AND Boolean Arithmetics
 

As i know when a client sends a query, the servers respond with files which match all of the words the search string has.

Example: Search String: gnutella pdf
Result: gnutellaprotocol0.4.pdf

This is not always good. If you wanted to find all the files containing the word gnutella and all the files containing the word pdf you should send two queries. One with the word gnutella and the other one with the word pdf. This kind of search could be done easily if Boolean Arithmetics supported by Gnutella Clients. This has nothing to do with the protocol.

The basic boolean arguments like AND, OR, NOT should be supported and maybe some extra. (For example, to make a search based on meta info). Also Case In/Sensitive searches could be implemented for more accurate results.

By using Boolean Arith. their is one more benefit. Almost every Gnutella Client has a small ListBox from where you can select the type of files you want to get as resuilts in a search. (Example: Images or Video, Sound etc.) What is really done in the background is that the client sends a query with your search string and when it receives the results, just filters them and you only see the file types you asked for. So your client gets, maybe, thousands results and finally shows you some hundreds of them. A more appropriate way to make such a search is to use (guess :D ) Boolean Arith. The User enter the keywords and when selects the file types he wants (let say Sound) the client adds some few keywords in the search string, in this example mp3 OR wav OR ogg NOT voc (which are too old, just an example of what you can do). By this way the traffic of Gnutella Network should decrease and earn more download speed and more accurate results.

Compatibility with older client (which do not support Booleans) it depends by which client sends the query :confused: . If an "old" client send a query with Boolean arguments inside a "new" client will understand them and answer with the appropriate way. But if a "new" client sends the query and an "old" client receive it, it will not understand the Boolean arguments (AND, OR, NOT) and try to find filenames containing these words. Probably zero! Or you maybe lucky and get some. The same thing happens if and "old" client send the query to also an "old" client.

Just a thought.... :D

very important
 

I think this is a very important topic as the major developers are considering adding flow control to limit the number of query responses. I think this kills one of the best features of gnutella and will make it nearly impossible to find what you want.

Instead they should reduce query traffic by making searches (not filters) more powerful. Over at the GDF in a related thread (http://groups.yahoo.com/group/the_gdf/message/5129) I suggested adding the following to the query:

-boolien operators (+, -, "", ect)
-file size parameters (min and max)
-file type
-meta data (bit-rate, resolution, codec, ect)

I would encourage everyone to let the developers know that limiting search results is unacceptable!!

Re: Gnutella AND Boolean Arithmetics
 

The AND search by default is a quick way to save bandwidth instead of returning a match for any of the words. With AND a search for "The ..." won't return too many results where in an OR search too many results would be returned so the servent would need to look at a stopwords file before sending.

For the second thing I quoted, for backwards compatibility a new message would probably need to be added that only contains the logical operators.

By searching for some wods that are ORed together is like making two or more seperate queries. If you wanted to search for three keywords that have no relation together ex: car, kitchen, garbage you should do three seperate searches for each word. But with boolean arith. you can simply type car OR kitchen OR garbage and have the same result. By doing this you have just saved some bytes because you send only one query instead of three. All gnutella clients can do multimple searches so it is up to the user to make them, you cant stop someone of making too many searches (and getting too many results also). Also the OR combined with the other boolean arguments provides a powerful way to search. Read the example above to my previous post.

I also agree with gnutellafan for using signs instead of AND OR NOT words and for the other search parameters.

By this way the Gnutella Network will become more "healthy" and avoid uselles search results.