Legend: SPEC!! - Not specified SPEC - Spec not finalized NICK - nick claims ARMA - arma claims - Not done * Top priority . Partially done o Done D Deferred X Abandoned For scalability: - Slightly smarter bandwidth management: use link capacity intelligently. - Handle full buffers without totally borking For dtor: NICK pre1: - make all ORs serve the directory too. - "AuthoritativeDir 1" for dirservers - non-authorative servers with dirport publish opt dircacheport - make clients read that and use it. - make clients able to read a normal dirport from non-trusted OR too - make ORs parse-and-keep the directory they pull down - authoritativedirservers should pull down directories from other authdirservers, to merge descriptors. - Have clients and dirservers preserve reputation info over reboots. - allow dirservers to serve running-router list separately. o "get /running-routers" will fetch just this. - actually make the clients use this sometimes. - distinguish directory-is-dirty from runninglist-is-dirty - ORs keep this too, and serve it - tor remembers descriptor-lists across reboots. - Packages define datadir as /var/lib/tor/. If no datadir is defined, then choose, make, and secure ~/.tor as datadir. - Contact info, pgp fingerprint, comments in router desc. - Add a ContactInfo line to torrc, which gets published in descriptor (as opt) - write tor version at the top of each log file pre2: - refer to things by key: - extend cells need ip:port:identitykeyhash. - also use this in intro points and rendezvous points, and hidserv descs. - figure out what to do about ip:port:differentkey ARMA - ORs connect on demand. attach circuits to new connections, keep create cells around somewhere, send destroy if fail. - nickname defaults to first piece of hostname - running-routers list refers to nickname if verified, else hash-base64'ed. pre3: - users can set their bandwidth, or we auto-detect it: - advertised bandwidth defaults to 10KB - advertised bandwidth is the min of max seen in each direction in the past N seconds. - not counting "local" connections - round detected bandwidth up to nearest 10KB - client software not upload descriptor until: - you've been running for an hour - it's sufficiently satisfied with its bandwidth - it decides it is reachable - start counting again if your IP ever changes. - never regenerate identity keys, for now. - you can set a bit for not-being-an-OR. - clients choose nodes proportional to advertised bandwidth - authdirserver includes descriptor and lists as running iff: - he can connect to you - he has successfully extended to you - he has sufficient mean-time-between-failures - add new "Middleman 1" config variable? - if torrc not found, exitpolicy reject *:* ongoing: . rename/rearrange functions for what file they're in - generalize our transport: add transport.c in preparation for http, airhook, etc transport. For September: NICK . Windows port o works as client - deal with pollhup / reached_eof on all platforms . robust as a client . works as server - can be configured - robust as a server . Usable as NT service - docs for building in win - installer - Docs - FAQ o overview of tor. how does it work, what's it do, pros and cons of using it, why should I use it, etc. - a howto tutorial with examples o tutorial: how to set up your own tor network - (need to not hardcode dirservers file in config.c) . correct, update, polish spec - document the exposed function api? - document what we mean by socks. NICK . packages . rpm - find a long-term rpm maintainer - code - better warn/info messages o let tor do resolves. o extend socks4 to do resolves? o make script to ask tor for resolves - tsocks - gather patches, submit to maintainer - intercept gethostbyname and others, do resolve via tor - redesign and thorough code revamp, with particular eye toward: - support half-open tcp connections - conn key rotation - other transports -- http, airhook - modular introduction mechanism - allow non-clique topology Other details and small and hard things: - tor should be able to have a pool of outgoing IP addresses that it is able to rotate through. (maybe) - tie into squid - buffer size pool, to let a few buffers grow huge or many buffers grow a bit - hidserv offerers shouldn't need to define a SocksPort - when the client fails to pick an intro point for a hidserv, it should refetch the hidserv desc. . should maybe make clients exit(1) when bad things happen? e.g. clock skew. - should retry exitpolicy end streams even if the end cell didn't resolve the address for you - Add '[...truncated]' or similar to truncated log entries (like the directory in connection_dir_process_inbuf()). . Make logs handle it better when writing to them fails. o Dirserver shouldn't put you in running-routers list if you haven't uploaded a descriptor recently . Refactor: add own routerinfo to routerlist. Right now, only router_get_by_nickname knows about 'this router', as a hack to get circuit_launch_new to do the right thing. . Scrubbing proxies - Find an smtp proxy? . Get socks4a support into Mozilla - Extend by hostname, not by IP. - Need a relay teardown cell, separate from one-way ends. - Make it harder to circumvent bandwidth caps: look at number of bytes sent across sockets, not number sent inside TLS stream. - fix router_get_by_* functions so they can get ourselves too, and audit everything to make sure rend and intro points are just as likely to be us as not. ***************************Future tasks:**************************** Rendezvous and hidden services: make it fast: - preemptively build and start rendezvous circs. - preemptively build n-1 hops of intro circs? - cannibalize general circs? make it reliable: - standby/hotswap/redundant services. - store stuff to disk? dirservers forget service descriptors when they restart; nodes offering hidden services forget their chosen intro points when they restart. make it robust: - auth mechanisms to let midpoint and bob selectively choose connection requests. make it scalable: - right now the hidserv store/lookup system is run by the dirservers; this won't scale. Tor scalability: Relax clique assumptions. Redesign how directories are handled. - Separate running-routers lookup from descriptor list lookup. - Resolve directory agreement somehow. - Cache directory on all servers. Find and remove bottlenecks - Address linear searches on e.g. circuit and connection lists. Reputation/memory system, so dirservers can measure people, and so other people can verify their measurements. - Need to measure via relay, so it's not distinguishable. Bandwidth-aware path selection. So people with T3's are picked more often than people with DSL. Reliability-aware node selection. So people who are stable are preferred for long-term circuits such as intro and rend circs, and general circs for irc, aim, ssh, etc. Let dissidents get to Tor servers via Tor users. ("Backbone model") Anonymity improvements: Is abandoning the circuit the only option when an extend fails, or can we do something without impacting anonymity too much? Is exiting from the middle of the circuit always a bad idea? Helper nodes. Decide how to use them to improve safety. DNS resolution: need to make tor support resolve requests. Need to write a script and an interface (including an extension to the socks protocol) so we can ask it to do resolve requests. Need to patch tsocks to intercept gethostbyname, else we'll continue leaking it. Improve path selection algorithms based on routing-zones paper. Be sure to start and end circuits in different ASs. Ideally, consider AS of source and destination -- maybe even enter and exit via nearby AS. Intermediate model, with some delays and mixing. Add defensive dropping regime? Make it more correct: Handle half-open connections: right now we don't support all TCP streams, at least according to the protocol. But we handle all that we've seen in the wild. Support IPv6. Efficiency/speed/robustness: Congestion control. Is our current design sufficient once we have heavy use? Need to measure and tweak, or maybe overhaul. Allow small cells and large cells on the same network? Cell buffering and resending. This will allow us to handle broken circuits as long as the endpoints don't break, plus will allow connection (tls session key) rotation. Implement Morphmix, so we can compare its behavior, complexity, etc. Use cpuworker for more heavy lifting. - Signing (and verifying) hidserv descriptors - Signing (and verifying) intro/rend requests - Signing (and verifying) router descriptors - Signing (and verifying) directories - Doing TLS handshake (this is very hard to separate out, though) Buffer size pool: allocate a maximum size for all buffers, not a maximum size for each buffer. So we don't have to give up as quickly (and kill the thickpipe!) when there's congestion. Exit node caching: tie into squid or other caching web proxy. Other transport. HTTP, udp, rdp, airhook, etc. May have to do our own link crypto, unless we can bully openssl into it. P2P Tor: Do all the scalability stuff above, first. Incentives to relay. Not so hard. Incentives to allow exit. Possibly quite hard. Sybil defenses without having a human bottleneck. How to gather random sample of nodes. How to handle nodelist recommendations. Consider incremental switches: a p2p tor with only 50 users has different anonymity properties than one with 10k users, and should be treated differently.