> I think this is needlessly complex. Simply add a "do not disturb me for X > seconds" to the QR reply. Then the receiving node will simply not route to > the rejecting node untill x seconds have passed. This will let overloaded
I'm sorry; but IMO this mearly leads to the accept queries for a fraction of a second; and lock down for a period of time situation that does happen already -- eg feast and famine situations; in the past; I had suggested that based on the number of querires rejected by a node in the past; factored by time to age out old things, that we use that to roll the dice against each time 'we would route' to that node to determine if we actually will route -- more rejections; the more we back off but not for n seconds; but rather by proportioning a number that we should back off by -- ideally our high volume peers would get to a state of backing off such that the QRs are down to a certain proportion -- if we can handle more; we QR less and they send us more; if we can handle less; we QR more; and we get fewer queries from them. No quotas ; targets ; decisons on 'how long will it take me to clean up this mess' and we don't have to maintain a list of who can ask me questions (I beleive we only have the list of who has asked me questions for diagnostics ...) I did take the time to write a minor change which put it into the probability to connect factor, and submitted it; however it was ignored. (I'll admit; it wasn't perfect; but I wanted to get the idea 'out there' to have it tried out before making it real clean -- I never got told if you make it cleaner we'll re-consider it; no-one was interested in talking about it; the 'powers' of freenet wanted to go in other directions; hense I packed up my things and went home -- I saw various comments re 'these people' that come in and try to 'contaminate' the technical ideals of freenet and then leave when it doesn't work really made me just go; heck -- you don't want to discuss an idea; for which I had the 'how to change the code to make the effect happen' -- why bother? Trevor _______________________________________________ Devl mailing list [EMAIL PROTECTED] http://dodo.freenetproject.org/cgi-bin/mailman/listinfo/devl
