Bearshare 1.3.0 rejecting Gnucleus?
 

A person using Gnucleus couldn't connect to my Bearshare to download a couple of files. I quit Bearshare and started-up Gnotella 1.0.0 and they are now able to download. What gives?

Sorry, I meant Bearshare 2.3.0

BearShare's response to a file request changed in its latest version.

Gnucleus had to be modified to accept this new version. It will be included in 1.4.2.

Wow! What did they do - add new header or something? I hope it's still HTTP compliant. Downloads are the only part of the gnutella protocol that actually follow a real standard after all, so I hope people keep with the HTTP specs when adding new features. I know in the past BearShare has done some wonky stuff protocol wise.

Since the evolution to 1.4.2.0 I have noticed a lot more instances of retrying hosts. in looking at the extended info, it seems that many of the connections to bearshare and limewire failed to connect. Is this something with gnucleus, or have bearshare and limewire changed something in their protocols?

If you are behind a firewall you will find many failures to connect since people who request a push are often also behind a firewall.

I test servers here and have not noticed any changes in any clients ability to download from gnucleus.

Hi Swabby! Have you noticed any decline in the number of search results lately? I'm seeing this in just about every peer, and am wondering if it's a general gnutella net thing (or perhaps some new peer not passing search results correctly).

It's especially noticeable searching for something rare (where there were never many results to begin with). I'd say on average I'm getting about half the results I used to. Looking at the statistics, the number of query packets to hits seems to reflect this too (it's about 35% to 5% for me now).

A decrease in the number of users of gnutella in general? I'm not sure what it could be, nothing to my knowledge has changed in how query results are passed back.

Since most clients now only reply to pings when they can take incoming connections, it is next to impossible to get a local host count.

I think there's probably more users than ever. I guess what concerned me was that maybe someone's peer has developed a bug (like dropping reply packets). It wouldn't have anything to do with Gnucleus, but could impact all peers. I just wonder how easy it would even be to track down something like that.

It's happened before - I remember almost a year ago BearShare introduced a TTL bug that flooded the entire g-net for a few weeks. Now there's new peers being marketed heavily (and a few somewhat dubiously) of late, based on code licensed from others. It makes me wonder if they understand the impact of any modifications they may make as they didn't write the original program. After all, this whole ball of wax pretty much depends on everyone playing by the rules.