On Tue, May 02, 2006 at 09:27:43PM -0500, Krishna Sankar wrote: > Nick, > > With due respect to you, I didn't get a good set of answers >for a couple of my question. Yep, I got the the general problem frame >- security and related artifacts. Thanks. Sorry for my vagueness, and thanks for your patience. > a) Is this still a good problem to work on ? i.e. do we have a >scaling pain point w.r.t. Hidden Descriptors ? If the upper limit is >~10K servers and we have only 100, it might be better for me to work >on some other tor topics, in the coding and design domain. It's not our current or even an immediate 'pain point' with hidden services as they are. Right now our biggest problem with hidden services is reliability and performance: connections seem to be slow, and go down often -- perhaps more than can be explained by the fact that they use longer circuits than typical Tor streams need. If somebody could investigate that, that would be a good start. > b) From my limited knowledge, hidden servers are trusted > servers and so we might be able to run SCTP between them as an > inter-server protocol. Hidden services aren't trusted; the directory servers that currently hold their descriptors *are*, but that's something we'd like to do away with. Right now, we have 3 trusted directory authorities that know about hidden service descriptors, and hundreds of directory caches -- scalability would probably suggest moving the info to other directory caches, not leaving it on the trusted authorities. > Going back to a) what other challenges are more pressing ? >4. Asybc DNS ? 6. Better buffer design ? 9. reverse DNS ? 6 and 9 are both good introductory tasks to get used to the code base; they are certainly smaller and more likely to get accepted into Tor quickly. I'm working on 4 now and should have something for it in a week or two. Are the others tasks that interest you? yrs, -- Nick Mathewson
Attachment:
pgpGGMiLEZmcw.pgp
Description: PGP signature