Ok here are some numbers. I think pretty representative.

I logged the network for 3 days 15h. every 2 1/2 minutes I generated searches for:

windows
exe
jpg
mp3
gif
madonna
american
avi

Alltogether 67 623 252 Hits. I had about 40 connections open in average (1GB LAN )

Here is the distributian and the percentages:
Fri Jan 11 19:54:21 2002
Tue Jan 15 11:17:38 2002
UPTIME: 3d 15h 23m
LIME:28246946 = 41.77%
BEAR:2718954 = 4.02% (strange???)
QTEL:20652 = 0.03%
CULT:0
GNOT:70665 = 0.10%
GNUC:1731415 = 2.56%
GNUT:45262 = 0.07%
GTKG:2860 = 0.00%
HSLG:0
MACT:671 = 0.00%
MNAP:576296 = 0.85%
NAPS:0
OCFG:0
OPRA:0
PHEX:136824 = 0.20 %
SNUT:29564887 = 43.72%
SWAP:0
TOAD:0
XOLO:216465 = 0.32%
ZIGA:105 = 0.00%
OTHE:4291560 = 6,35%

very interesting! I wonder why bearshare is so low? Maybe vinnie's new version already bans other clients.
What a dork. I'm a little surprised to see swapnut with so many.
What client did you use for this little experiment?

My hacked Qtella client

How can you call it representative, if you used Queries that modern clients do not respond to purposedly ("exe", "mp3", "gif", "avi"). Hence your low turnout for some popular clients, such as BearShare.

Other than that, you're searching for rather specific things ("windows", "madonna"). How about using more broad keywords like "live", "metal", "sex", etc.

Another thing: You're purposedly are spamming the network every 2-1/2 minutes. In my opinion, I cannot call it reasearch if you disturb the network.

-- Mike

Calm down a little bit ... I mean I do that for 3 days that's it. By the way you know how many clients of the network you reach with the searches? Not as many as yo expect

Do you have a better idea how to check out the client information? At last I have some very interesting results about network size and structure from it and as it looks like Qtella would also scale if no TTL limit would be used. Is that research enough? Shure I do generate load on the network you are right but how would you like to improve the network concept without any fundamental research and theory on the network itsself?

Just to make it clear, I do not do that for fun and for months. I started it last week and next week the messurements are done. AND the client isn't always running as well ...

Seams all clients that I know respond to these queries. But you got a good point I kicked them out. I added "live", "metal", "soundtrack" and "sex" to the queries. Some more suggestions?

Interestingly they should block a three letter search for "exe" but they find "sex"

The reason why so many people have voted for Xolox (and are using it) is because it actually WORKS. I have been systematically going through Gnutella clients trying to find something decent to use and NONE of the up-to-date clients can match the results of Xolox.

If you are going to block Xolox then perhaps you should also block all of the people using older versions of Limewire and Bearshare (since everybody knows the new versions are complete sh!te). Hell, lets go one step further and completely kill the network by blocking anyone that isn't using THEE latest client implementing THEE latest non-working features!

Whatever...

Anyone who implements blocking of specific clients should perhaps think about starting their own bloody network.

Hi hermaf, cultiv8r!

Hmm about client statistics, we have three posibilities:

- high TTL search Queries and evaluate vendor codes from the 'Extended Queryhit Descriptor'.
- share common request files and evaluate HTTP log files.
- a low TTL crawler (TTL=2 or 3) and evaluate handshake... ONLY with v0.6 servents.

Unfourtunately the last possibility does not work with older clients (e.g Xolox). But I guess it will give us a much better client statistics in future! Such a crawler is not unhealthy (no excessive broadcast traffic), results do not rely on shared files and routed messages, are not faisified by pong caches, throtelling or file caches in superpeers.

We could combine all three methods, compare the results.

/Moak

[QUOTE]Originally posted by Moak
I have analysed Bearshare and Xolox a while.

Comparing some of the most favourite clients, Limewire, Xolox, Limewire, Phex, Bearshare....
Bearshare is those with less functionality, less technology, the oldest in this context!


Could u plz. supply me with this analysis?? It will help me so so so much in the process of deciding how to make my client a healthy, good behaving client

I'm too lazy, just list all the features a client has and compare.

Why blocking clients that are no longer being developed??? Would you take all old Beetles from the streets just because they are not being produced any more? What about the people that still want to drive them?

I'm currently using Gnotella with no problems which development is (momentanously???) stopped. I am not comparing Gnotella with the Beetle, but I know there's still a lot of people using it, and I know it's more than the 0.10% mentioned above. Besides, those satistics are relative. Some clients are used by some specifically for some types of files, and therefore the words you use to compare may not be the most adequate to do an objective analysis of the network.

Ivan
"In the dark we make a brighter light"