From: Mike A. <mi...@yu...> - 2001-08-29 21:52:14
|
Okay, If people aren't on the GDF but are talking about standards I suggest they join it. th...@gr.... Some worrying things about setting standards is getting to agree them, and the GDF is very good at not doing that. So far many different things have been mentioned on there and come to nothing. So I am going to say some words about some of the topics recently discussed. I also aim to present a web page with all the relevent links and discussions on, a work in progress. FFH: or Full-File-Hash. Basically the 'is this file the same as all these called the same thing. The suggestion was to include a null in the query with "##" then another null. This would signal the 'I want a hash' flag before XML provides a method. (More on XML later). The servents would then respond with hashes if supported. The digest methods agreed (in it's loosest sense) were SHA1 and TigerTree as a compound element. A few options were discussed about partial support though. (Which made me happy as Python has SHA1 built-in, for my own client model). QRP: Query Routing Protocol. The method to direct queries to servents with files shared that share the same elements of name. Quite involved passing mechanism and 'hashed' word table generation (I can't see what is wrong with plain text...). It works out very good for freeloaders as no query strings get sent to them. It also works very well for specialist interests people as the number of 'porn' and general queries are reduced. So if I had a hard disk full of Mike Oldfield tracks, I would not get queries for Porn, DivX, Maddona etc. Connection Negotiation Handling: Well the first major agreement I can remember the GDF actioning is the protocol handshake change. Extensions are being discussed all the time, and no method of negotiation exists presently. The new protocol is now: > Gnutella Connect/0.6/r/n > <RFC822 and Extended Headers>/r/n/r/n < Gnutella OK/0.6/r/n < <RFC822 and Extended Headers>/r/n/r/n The headers have not been agreed as of yet. Download Queue Mechanism: Been discussed, no article written as yet... <watch this space> XML Metadata: The metadata question was raised and discussed. A basic overall agreement was to use XML, but side discussions about alternative ways to kill XML via tokenising, compression just for XML etc were discussed. No single agreement has appeared to my knowledge. Connection Compression: There was a discussion about compressing the connections using Zlib. People couldn't agree (as is typical on the GDF) as to whether that was the best method. (As Zlib is available on most platforms for free it sort of stands to reason). This was going to be toggled somehow by the new connection method. They are the only ones I can think of presently, I will try and get the links on a single page soon, but in the mean-time the GDF does have an archive, but not as mbox :(. Thanks Mike Albon |