resume = multiple sources possible??
 

If bearshare can resume a file that is half way completed, there must be a way we can download from multiple users (pretending to be resuming from different positions). The same way KaZaA (i guess) uses. 1.6Mbit/sec sure is appealling...

If it's impossible over the gnutella network somebody tell me why.

Neither impossible, nor difficult...
 

Your servent, whichever you use, obtains the files by sending an HTTP-format GET request. The parameters of that request include the range of bytes you want from the file. That is how resuming is possible; you simply request the range of bytes you don't already have, and append it to your (partial) file.

More complicated servents can and do request different ranges from different hosts simultaneously. Doing this shared a requirement with resuming; you must ascertain that the file being shared is the exact same file you already have a part of.

But there's nothing preventing multiple source downloads.

Since most Gnutella servents have bigger fish to fry, I prefer to just start two downloads of different material from different sources; they'll finish at the same time they would have using the other method, as I am narrowband (therefore using two sources guarantees that my bandwidth is always full). You may want to use five...

multi source
 

Then a client should do it!!! There are 2 things I don't like about gnutella.

1) I was looking for bagpipes music the other days and tried something like 30 transfers yet none worked. I went in another program (wich doesn't use gnutella net) and got 2 mp3s right away.

that, we cannot do anyhing but

2) top speed around 200kbps while multi source downloads reach 1.6Mbps often

exact same file is easy to see: when the file is "gnutella forum movie part 1_portchop.avi" and 368520211kb" another file just could not have same name & size!! maybe the program could just prompt me and ask :
"Is this the same file and would you like it to be added to the download as another source or something?"

If there are any programmers working for a gnutella client out there please tell me this will be done! I can only program SOME basic (yep, old gw-basic) so I guess it is a lot more complicated to add this kind of feature to a program but still I don't think it would take 40hours...especially when there's allready a resume feature in the program!

Re: Neither impossible, nor difficult...
 

To be sure that the two files (or more) are the same is easy by checking the checksum of the file. Thus, making a gnutella client download from multiple users should only check the file sizes and the crc (checksum) thus making it possible to download files that are identical but with different filenames from many user at the same time.

By splitting the file into say 100 parts the client can download one part per user from all the users that have the same file. As soon as one part is done from one user, a new request is sent to that user.

To make the Gnutella network competetible also the supernode functionability (like fasttrack based program uses) need to be implemented. The supernode keeps track of all the files (metadata, including checksums) that the connected clients are sharing, making the search 10 times faster and increasing the maximum network size.

Use XoloX Gnutella client. it can download from multiple hosts giving you 300k/s plus speeds.

plus it has some other features the other clients can only dream about :)

www.xolox.nl. or use client forum on this forum.