On Wed, Jun 21, 2006 at 07:45:03PM -0700, Ian Clarke wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
> 
> 
> On 21 Jun 2006, at 09:11, Matthew Toseland wrote:
> 
> >On Tue, Jun 20, 2006 at 09:27:43PM -0700, Ian Clarke wrote:
> >>-----BEGIN PGP SIGNED MESSAGE-----
> >>Hash: SHA1
> >>
> >>Still thinking about this - but is there any place here for the
> >>"minimum request interval" mechanism we used in 0.5?
> >
> >This supercedes the minimum request interval mechanism, by passing
> >tokens rather than trying to determine rates. It's better - it's more
> >precise, it reacts more accurately as well as faster. It more  
> >accurately
> >simulates the underlying metaphor or incompressible fluid flow in a
> >network of pipes.
> 
> One of the points of the MRI mechanism was to reallocate (what are  
> now being called) tokens from nodes where they weren't being used to  
> nodes where they were.
> 
> In the proposed scheme, what happens if one neighbor is sending a lot  
> of requests, but none of the other neighbors are sending any, so the  
> one "greedy" neighbor's needs can be satisfied?

Tokens are allocated either fairly (add the token to the emptiest
bucket) or preferentially to nodes which use less requests (add the
token to the fullest non-full bucket)). If there is sufficient capacity
then all requests will be accepted; if not they will either be allocated
fairly or they will be allocated first to nodes sending few requests and
then to nodes sending many requests.

The great thing about this is that it's really hard to propagate MRIs.
Whereas if we have specific tokens, it's relatively easy to propagate
them.
> 
> Ian.
-- 
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/20060622/005ccdc6/attachment.pgp>

Reply via email to