Hi !

I have to admit, that I caused many of the threading problems all by 
myself <g>
Unfortunatly requests from clients aren't rejected when the node gets 
overloaded.
And thus by telling frost too start many downloads or update many 
boards, a (slow) node
can be killed quite easy :(

I think while this is not as important as not accepting too many 
requests from the
outside world, also the fcp requests should be limited.

I see currently two possibilities:

1.) Extending the protocol to include the possibility to temporarily 
reject requests.
This would writing clients much more complicated and therefor (imho) 
isn't a good idea.
Also clients cannot possibly know, how much load a node can handle and when
to start the request again.

2.) Delaying the requests until the load is lower again. Which is nicer 
(imho),
but currently I do not see where this fits in the current implementation.

Niklas


_______________________________________________
Devl mailing list
Devl at freenetproject.org
http://lists.freenetproject.org/mailman/listinfo/devl

Reply via email to