On Saturday 26 January 2008 13:08, Michael Rogers wrote: > Matthew Toseland wrote: > > - Some form of per-node failure tables (requires modifications to ULPRs). > > Just a quick thought: any idea how per-node failure tables would affect > opennet destination sampling? If a successful request follows several > failures (possibly from other requesters, so the successful requester > won't be able to tell) it seems like it could take quite a convoluted > path...
The node at the end should be close to optimal even if we've redirected near the beginning, right? Otherwise, we've found data that was somewhere it shouldn't have been, which is valuable because it will then move to where it should be, although we get a dubious route in the short term - it shouldn't happen very often? If finding data in odd places is very common, then path folding from ordinary routing isn't going to work very well either, it'll be dominated by fetches of ubiquitously cached data (which it probably is...). > > Cheers, > Michael -------------- next part -------------- A non-text attachment was scrubbed... Name: not available Type: application/pgp-signature Size: 189 bytes Desc: not available URL: <https://emu.freenetproject.org/pipermail/devl/attachments/20080126/e0083704/attachment.pgp>
