[Tech] Re: Re: Backoff considered harmful?

[Michael Rogers]

Jano wrote:
> Newer stuff completes instead of nothing completing? A FIFO queue can have
> unbounded latency, while a LIFO has just the servicing latency for the
> requests that manage to get a slot.

Interesting idea - at the moment I'm trying to model the behaviour of 
the current network in order to evaluate token passing, but once that's 
finished I wouldn't mind taking a look at this - although as Toad 
pointed out the queue lengths should be controlled by token passing anyway.

The queueing logic would have to be a bit more complicated if we were 
going to drop messages under heavy load, because it would make sense to 
give priority to messages that belong to existing searches/transfers 
rather than starting new searches/transfers and allowing the existing 
ones to fail... still it's something to think about.

Cheers,
Michael