Bill Pringlemeir wrote:
> On 5 Nov 2006, [EMAIL PROTECTED] wrote:
> > Well, as said, there's the overlap check. /dev/urandom from a single
> > source would work but not from multiple - assuming they're not using
> > the same /dev/urandom or there's at least one good source. I would
> > think if you cranked the overlap check size up to the maximum, you
> > should eventually get the correct file especially with such a huge
> > amount of sources.
 
> It is rather easy with a congruent random number generator.  See
> Knuth, et. al.  The SHA1 is a fine seed value.  If overlap ranges are
> known and the complete file is available, then feeding valid data at
> the start and end of ranges is also not a problem.  Many files only
> need a few bytes in a header screwed up to make them unusable; only
> changing one byte in a middle of the range is needed.

Nonetheless it might help if we increased the overlap size on a mismatch.  At
least I've noticed that with a larger size, it's not a problem to download such
files even if there are a couple of mismatches.

On a related note, I've also added a property/setting to configure the maximum
number of downloads per file. So you won't see this ridiculous swarming with
1000 sources for a file that's just a few megabytes large or even smaller. The
default is 20 which should be fine for most downloads. I would think that also
decreases the chances that you get a bad sources into your swarm since they can
only serve so much.

Further, this should also allow more parallel downloads of multiple files as a
single file won't consume all of your download slots.

-- 
Christian

-------------------------------------------------------------------------
Using Tomcat but need to do more? Need to support web services, security?
Get stuff done quickly with pre-integrated technology to make your job easier
Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642
_______________________________________________
Gtk-gnutella-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/gtk-gnutella-devel

Reply via email to