GUESS in 2003 or 2004?
 

Wow,
Is it just me or is it taking for ever to wait for the much hyped GUESS? I hope its worth the wait. With all due respect to the developers, advancing the gnutella network at such a slow rate is not going to sufficiently win people over. Fasttrack has been out for over a yr and gnutella still pales in comparison. I agree with the Mike at Shareaza for doing something bold. I m sure GUESS will improve search results but what is still missing is the scalability to search the whole network. Maybe that is the next step and is in the works. Nobody wants to wait yrs just to get a network that performs as well as those several yrs past. I dont mean to come off as abrasive but this is how I feel.

However, I do appreciate the effort by the LW developers and others.
Any projected release dates on 3.0?

3.0 will probably be released next month. There are still a couple of projects that are not complete, like using 19 instead of 6 inner ultrapeer connections (while reducing ttl from 6 to 4), using qrp for ultrapeer connections, improved firewall detection, vendor messages. - I'm not sure about magnet links will be ready in the next version... but I hope so. The GUESS implementation is already working in the development version (well it has been working more or less for about two months now).

Hm, wouldn't it be good then to release a final paper for the developers so that everyone can start working?

Latest news is, that UDP will not be used for searches despite LimeWires prior announcements and their efforts to implement GUESS. The latest development ultrapeers have 42 intra-ultrapeer connections, TTL is reduced to 4 and query-requests will be sent to one of the ultrapeer connections at a time so LimeWire will be able to stop the query once it has a certain number of results. GUESS won't allow such an extensive use of QRP as this scheme does and searching 50,000 ultrapeers with GUESS would be a pain while it isn't a problem for a TCP searches.

Thats a joke i think? All the hope i layed into GUESS just vanished away :(
How will youimplement global searching without UDP!?

Tell me that this was a joke, please!

That is not a joke. UDP for searches is more trouble then it seems. The improved TCP searching will allow LimeWire to search about 50,000 ultrapeers according to Greg Bildson (the gnutella network currently has no more than about 20,000 ultrapeers). In addition it will allow the servent to make use of QRP between ultrapeers and flow-control if necessary,

GUESS as it was had the problem that it was relying on the pongs it was receiving via TCP and together with querykeys there was not much bandwidth improvement over TCP searches. I believe for global urn searches CHORD or KADEMLIA (or even an improved version of current download meshes) could work much better than GUESS and I don't believe we need global keyword searches.

I admit I was a little disappointed GUESS would not be implemented, but when I think about it now, selective querying will work much better.

heh...so GUESS isn't even coming out?

Guess might be only for requieries at first, we'll how selective searches work in the wild first.

LW better live up to expectations
 

OK.
The wait is really long. I hope for LWs sake they bring to the table some type of network improvement that finally brings gnutella up to the performance of fasttrack. I ve been a patient LW fan for yrs but I cant wait my whole life to have a decent functioning network. I may have to abandon ship if things dont significantly improve quickly. Gnutella needs to be a leader and not a follower in p2p technology.

I agree completely. I've not been using LimeWire that much lately because Gnutella is in such a pitiful state.

i really hope LW improves there network so gnutella will improve. it seems like Shareaza is the only one doing something to improve gnutella with G2 (but others have their own negative opinions about G2). i think it helps the network, although some think it hurts it.

don't be concerned that GUESS isn't being put to heavy use. LimeWire has still put support for it in the software, and may plan to use it for requeries... so if future versions decide to always use it, old versions will support it.

also, the fact that they're not using GUESS exclusively says alot. it means that despite being able to attribute it to themselves, they realize that there are other (better) ways of improving the network. in this instance, the better way is 'dynamic querying' ... see http://www.gnutellaforums.com/showth...5107#post65107 for a short overview of how it'll work.

I don't think using GUESS for requeries is such a good idea. Since requeries should return only very few results, so the benefit of sending queryhits directly is very small. Besides, in total, less query messages have to be sent using dynamic searches to search the same number of ultrapeers thanks to QRP between ultrapeers. Don't forget that using GUESS, intra-ultrapeer QRP is pretty useless.

You should really test GUESS a little (you will have to disable the QueryKey for that). You will find that tcp queries come a lot closer to a global search than GUESS, because in one hour you will receive merely enough pongs to search about 1,000 ultrapeers and I found that 30-40% of them do not even answer to the query because the packet was either lost, those hosts are firewalled or not even online anymore.
The amount of Pongs that would have to be sent to allow using GUESS for global queries would be huge and there are already more pongs than many ultrapeers can handle. They have to devote 20% of their upstream for forwarding pong traffic alone.

since the ttl was reduced, my search results have taken a nosedive. Searches that returned 30-60 results before get me 1-10 now. News that GUESS won't happen, and that it wouldn't have done too much anyway, makes me wonder if I'll ever get results again.

I'd always wanted to keep going until 3.0 at least, think I may just bag it instead. blah. Someone change my mind fast

The key lies in the reduction of message traffic with intra-ultrapeer qrp & dynamic querying.

At the moment most of the ultrapeers are simply overloaded and they have to drop some of the messages they receive. The result is, that a query that might have reached some 50,000 ultrapeers (= about 1,000,000 unique leafs) will not even reach 3,000 because it will be dropped before that. So you see, in theory, tcp searches could easily search more than three times as many hosts as there currently are on the network. Practically your queries will be dropped so often that you can only search a fraction of the existing network.

The explanation is simple:
* There are far too many requeries, - even LimeWire clustering can't shield against all of them.
* Pong traffic requires lots of bandwidth, although it does serve a purpose.
* Searches for popular content return easily more than a thousand results and many results are probably already dropped before they reach you in that case.
* LimeWire limits the bandwidth allocated for gnutella traffic very strictly.

The hope is, that LimeWire 3.0 will change it all and make gnutella as good as Fasttrack and I am confident that it will manage to improve the network significantly, so it cannot be compared to the old network anymore.

any feedback on G2 perfomance?
 

Has any made any observations on G2s performance? If it lives up to what Mike says it can do perhaps it would be an alternative solution. Supposedly you are able to search the whole network.

G2 has yet to prove itself. There are only a few hundred G2 hubs and with such a small network even regular gnutella worked very well. Maybe somebody could tell if the protocol was published but we are still waiting for the specs to be released.