LOL. Thats what i was gonna say, but I wasnt sure exactly how it worked. I was gonna use a nice VW Bug vs Mack Truck analogy, to explain it better.

I didnt even -think- about packet loss due to bad hardware, connections. I kept equating distance ::LOL::.

Thx again for clarifying that.

I don't know, but I think it was Napster (god please bring back napster) that showed ping and I always went for those with lower ping, it's usually the faster ones. I just did a search that turned out 263 results, there was ONLY 5 cable modem users, 37 T1, and the rest ALL T3... something's wrong in this picture... if you can only display user-selected connection types, you might as well erease that field.. it's useless, untrue information.

How bout autodetecting speed?

Doing an upload and download bandwith test during the install, no manual selection.

Then, if a user tries to block the speed test, or doesn't want to perform it, dont install.

A user can stop all internet activities and install so he gets a near-perfect performance mark for his connection type. but who knows how many other sharing program he's running or how many files he's downloading at the time you are doing the search?

how about a dynamic autospeed test? the test could test say once every hour and then the speed would change according to what the test results were.

Sounds good. Small file test, though. Mebbe 100k. To not screw with the 56k'ers.

What file would you want to download? And where from? And how would you make sure it's the local host causing a lower speed and not the remote host?

D/l from a HTTP server, of course. Not from a remote host.


(edit: by "not a remote host" I meant not from any old gnutella user.)

A HTTP server is a remote host. And who do you think is willing to put up a server for that? Something like that could easily generate a few GB of traffic per hour...

True.

There has to be a way that we can accurately measure throughput speed without needing a special server to hammer.

How bout.... when the user downloads files, it takes the speed of the d/l every few second(10kb/s, 20kb/s, 15kb/s...) and then averages it. It then stores it in a database. When the user gets an upload request, it averages the upload the same way.

Then, the avg. upload & download speeds, max recorded upload & download speeds could be reported when you hover over the person with your mouse to d/l.

(hover tooltip)
-------------------------------------------
| Client Name_.: Default_______.| (1)
| Avg/Max UL__: 25/31 KB/s____.|
| Avg/Max DL__: 43/55 KB/s____.|
| Files Shared_.: 156 (405MB)___| (2)
| Curent ULs__.: 1 (2 slots left)__| (3)
| Est. Wait____: 0 sec_________| (4)
| IP Address__.: 121.56.16.xxx__| (5)
-------------------------------------------

1.) Let clients name themselves. Mainly just to have names in the chat window.
2.) Let users know how many files a person is sharing (amount and overall size)
3.) Let users know how many files a person is currently uploading, and number of slots left.
4.) Estimated wait time before downloading a file (queue length, basically)
5.) IP address, masking last octet.

Edit: Screw Ping Times. I thought it was a "might as well" sort of thing. But they -are- useless, after all.