On Sun, Nov 10, 2002 at 01:27:45AM +0100, Oskar Sandberg wrote:
> 
> Please don't advocate flooding the network based on anecdotes like this. 
> If people respond by flooding, then things will get worse, and we will 
> never know if they could have gotten better.
A small number of retries often pulls the file through. This does not
produce very much load, compared to the great reduction in future hops
caused by failures. 1000 retries for 1000 files would be a problem.
Maybe fish should limit the number of retries. Besides, users will
always take reasonable steps to get the file, and tools that use a few
retries, or even use parallel requests, will be popular. See FEC encoded
splitfiles.
> 
> Can people please show some restraint in regard to drastic action to 
> make things work better. The network has not been stable long enough to 
> draw any conclusions about how well it works as is.
> 
> 
> On Sat, Nov 09, 2002 at 11:23:24PM +0000, Matthew Toseland wrote:
> > Recently I inserted a file at HTL 20 into one of my nodes, and had it
> > DNF on the other at HTL 25. For two different files. This worked in the
> > past, with lower HTLs. Now, the interesting part is that I have three
> > nodes, and it appears that hammering all the nodes other than the one I
> > inserted the files into for all the files improves fetchability over
> > time... maybe fishtools etc should include an option to fetch from extra
> > nodes? The modified loop:
> > Insert files.
> > For each file:
> >  if fetchable on all nodes, continue to next file
> >  for each extra node, (in parallel?):
> >   try to fetch at current HTL
> >   if RNF, retry with same options
> >   if DNF, and HTL < max HTL, increase HTL and retry
> >  done
> >  repeat
> > done
> > 
> > Eventually the files should become available... and the network will
> > forge new links to make this happen. The only cost is time and load...
> 
> -- 
> 
> Oskar Sandberg
> oskar at freenetproject.org
> 

-- 
Matthew Toseland
toad at amphibian.dyndns.org
amphibian at users.sourceforge.net
Freenet/Coldstore open source hacker.
Employed full time by Freenet Project Inc.
http://freenetproject.org/
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: not available
URL: 
<https://emu.freenetproject.org/pipermail/devl/attachments/20021112/572938ef/attachment.pgp>

Reply via email to