[freenet-dev] [freenet-cvs] r11679 - in trunk/apps/load-balancing-sims/phase7/sim: . clients messages

[Michael Rogers]

Matthew Toseland wrote:
> Simulations maybe?

You could find simulation parameters (link speed, latency, reordering) 
where less than 5% of the retransmissions would be unnecessary, and 
other parameters where more than 5% would be unnecessary. Without an 
adaptive mechanism it's hard to know which parameters are realistic. 
That's why the reordering paper threw me for a loop - I'd assumed we 
could just pick a single realistic value, but some links seem to reorder 
much more than others.

> What exactly does TCP do? I thought it combined ack's with data in order
> to minimize overhead.

Right - data will be piggybacked on an ack if there's any data waiting 
(and the same's true for the current simulation code). But if there's no 
data waiting the ack will be sent on its own.

Cheers,
Michael