Talk:Comparison of Gnutella software

Removed items without information
There are many clients without information in the table. It makes the table less useful to someone actually reading it. It may have been nice to have a large list if you were editing to add them all. However, I think the Wikipedia reader should come first. Bpringlemeir 23:13, 25 April 2007 (UTC)

Useless entries?
Some of the table entries are no longer relevant. For instance, GWebCache is no longer a preferred method of bootstrapping. The GWebCache wasn't very scalable. Also, there are several columns dedicated to router issues (UPNP, NAT traversal, NAT port mapping and F2F). That makes the table look weighted toward these, but they really aren't that important for a Gnutella program to function properly. I think that newer features like dynamic query, high out degree, and tiger tree hashing would be more relevant. Also, the memory and CPU requirements would be a very nice criteria to have. But I guess that is sort of an idealist. It would be difficult to get this information (and keep it consistent across versions). Bpringlemeir 23:22, 25 April 2007 (UTC)


 * I disagree, because usability behind a router is very important especially on the Windows platform as most users have trouble setting up port forwarding manually. Gnutella does still work even if the machine appears to be non-connectible from the outside but the network overall suffers because these machines cannot operate as ultrapeers and they cannot upload to any of their own kind either - unless they can use LimeWire's NAT-to-NAT protocol. That said, I understand what UPnP support means but I don't grasp what 'NAT traversal' and 'NAT port mapping' are supposed to indicate in this context. If the latter refers to NAT-PMP I would think that's part of UPnP, even if not this could probably me merged in the table, especially as everyone seems to support either both or none. I'm almost afraid 'NAT traversal' refers to SOCKS? If so, that should be indeed fairly irrelevant and rumour has it those who do support this, don't support it well at all - not that SOCKS is wide-spread anyway.
 * Resource requirements heavily depend on how people use software and also on the specific environment. Also some may use little resources initially but the usage might not scale well, whereas some may use more but have a more scalable use of resources. Likewise those that use little resources might implement very little of Gnutella (correctly). Most non-LimeWire-derivates don't have Unicode support at all which means they are not only misrouting queries, they also save some amount of memory and CPU - something which many English-speaking users might not care about or be aware of. Last but not least, a client which supports only leaf mode will of course use very little resources compared to a software running in ultrapeer mode. Thus such a comparision would not only be difficult it might also send a false signal and give a wrong idea about the efficiency of the software. --82.141.49.144 13:47, 16 May 2007 (UTC)


 * Ok. I don't think you actually disagree.  I wasn't saying that router functionality weren't important to some people (ie, Windows users).  I was saying that it wasn't a Gnutella network feature and it is a significant section of the current table.  I was thinking that it should be in a separate table or a single entry with the type of router support.  People using a Unix operating system could ignore that table or entry.


 * gtk-gnutella definitely supports Unicode and it is more efficient than a Java implementation. Phex also supports Unicode afaik.  This is an excellent point, Unicode should be in the list.  Oh, it is.  This is more important than GWebCache support for instance.  I agree that performance and memory usage could be difficult to quantify. gtk-gnutella for instance has different performance using gtk1 or gtk2.  I did think it would be a nice entry for someone searching for a client to have, but it might be impossible to provide.  Bpringlemeir 21:01, 18 May 2007 (UTC)


 * Maybe the table should be split into core features and maybe convenience features. "Chat", "Buddy List" and the whole NAT stuff has little to nothing to do with Gnutella itself, the first two are completely orthogonal. For the record, "Unicode" is core features, it has nothing to do with Unicode support in the GUI. I'd also add a TLS column, as LimeWire Beta supports this as well but the table is already rather wide. Some of the older software does not support Out-of-Band querying as far as I know which makes them less useful and/or puts more strain on the network. --22:03, 2 August 2007 (UTC)

Useless clients?
There are several clients that had no information and are never seen on Gnutella. I believe that most people might consult this page to get an idea of what client might be for them. Acqlite, Cabos, and CocoGnut are not currently very popular. The clients Apollon, BearFlix, DM2, Gluz, KCeasy, Kiwi Alpha, Poisoned, Swapper.NET, XFactor, and XNap had no information, but I don't even know if they are worth mentioning. Perhaps some sort of column for a date of last release would give an idea of how actively maintained these projects are. If a client hasn't been updated in the last three years, it would not connect to the current network. Bpringlemeir 23:29, 25 April 2007 (UTC)


 * Many of the clients I mentioned use the giFT back end and only supply giFT as a name. Bpringlemeir 00:17, 10 May 2007 (UTC)