That's how it works now.
-- David
[email protected] wrote:
> My apologies for the late post, I was on vacation for a couple of weeks.
>
> Adaptive replication should never completely trust a computer, there should
> always be some chance that it will get a wingmate for any particular
> result, even if it is 1% or so. A single failure noted by random check
> should instantly make that computer untrusted. It is my belief that most
> computers would be trusted, and therefore any untrusted computer would be
> likely to get a trusted computer as its wingmate.
>
> My thought would be that every time a trusted computer is the first to get
> a result, a random number is generated and if it is less than some
> threshold, the computer gets a wingmate. Of course, a completely untrusted
> computer would have a threshold of 1, but even the most trusted computer
> would have a threshold of something like 0.01.
_______________________________________________
boinc_dev mailing list
[email protected]
http://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev
To unsubscribe, visit the above URL and
(near bottom of page) enter your email address.
