On Thu, Apr 13, 2006 at 12:46:41PM +0100, Michael Rogers wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
> 
> Sounds good, but could you clarify a couple of points?
> 
> The "request send rate" applies to forwarded requests as well as
> originated requests, right?

Yep. One request send rate for all neighbours combined.

> And there's a separate rate for requests
> sent/forwarded to each neighbour (as opposed to one rate for all
> neighbours, or one rate per request)?

No. One rate overall.
> 
> When a RejectedOverload is sent to a neighbour, does the neighbour
> reduce its send rate using AIMD, backoff or both?

If it's local, it backs off. Either way it reduces its overall send
rate via AIMD.
> 
> If the neighbour wasn't the originator of the rejected request, does it
> forward the RejectedOverload back towards the originator, and if so, do
> all nodes along the path reduce their send rates using AIMD, backoff or
> both?

Yes, that's the idea. A RejectedOverload or a timeout on any request
reduces the global window (or the per-type window); a non-timeouted
request increases it.
> 
> Also a suggestion: could RejectedOverloads be sent immediately based on
> the queue length, rather than waiting for 10 seconds?

We could. We could alternatively use some form of simulated queueing to
avoid queue latency, although I'm not sure offhand how this could be done
fairly; I'm sure we could devise a fair simulated queueing algorithm.

So no additional latency either.
> 
> Thanks,
> Michael
-- 
Matthew J Toseland - toad at amphibian.dyndns.org
Freenet Project Official Codemonkey - http://freenetproject.org/
ICTHUS - Nothing is impossible. Our Boss says so.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 189 bytes
Desc: Digital signature
URL: 
<https://emu.freenetproject.org/pipermail/devl/attachments/20060413/ddc44d83/attachment.pgp>

Reply via email to