TODO 14 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328
  1. Legend:
  2. SPEC!! - Not specified
  3. SPEC - Spec not finalized
  4. NICK - nick claims
  5. ARMA - arma claims
  6. - Not done
  7. * Top priority
  8. . Partially done
  9. o Done
  10. D Deferred
  11. X Abandoned
  12. 0.0.8:
  13. NICK - Why is the first entry of kill -USR1 a router with a 0 key?
  14. X why did common/util.c build-depend on or/or.h ?
  15. NICK? - Tors deal appropriately when a newly-verified router has the
  16. same nickname as another router they know about
  17. o Add '[...truncated]' or similar to truncated log entries (like the directory
  18. in connection_dir_process_inbuf()).
  19. NICK . put ip:port:keyhash in intro points, rendezvous points,
  20. and hidserv descriptors.
  21. o Specify
  22. - Implement parsing
  23. - Generate new formats (Not till 007 is dead)
  24. NICK - unify similar config entries that need to be split. put them
  25. into a smartlist, and have things take a smartlist.
  26. o if a begin failed due to exit policy, but we believe the IP
  27. should have been allowed, switch that router to exitpolicy
  28. reject *:* until we get our next directory.
  29. o tor-resolve needs a man page
  30. - "AllowUnverifiedRouters" config option
  31. NICK - Parse it into 3 bits
  32. ARMA - Consider it when picking nodes for your path
  33. ARMA - if there's only one entrynode preference and multiple exit node
  34. choices, don't pick the desired entrynode as exit.
  35. o 'fascistfirewall' option to pick dirservers on port 80 and
  36. ORs on port 443.
  37. NICK - extend it to take a range of ports
  38. ARMA - don't replace options->Address with the resolved one at startup.
  39. keep the old one and also have a new one.
  40. ARMA - detect our address right before we make a routerinfo each time?
  41. ARMA - external IP vs bind-IP
  42. NICK - parse uptime into router->uptime
  43. o clients choose nodes proportional to advertised bandwidth
  44. o and/or while avoiding unreliable nodes, depending on goals
  45. - figure out what to do when somebody asks to extend to
  46. ip:port:differentkey
  47. * reject it. assuming this is as dumb as it sounds.
  48. - make loglevel info less noisy
  49. o defining an ORPort isn't necessary anymore, if you use
  50. ORAddress:port. Same with DirPort, SocksPort.
  51. bug fixes, might be handy:
  52. - the directory servers complain a lot about people using the
  53. old key. does 0.0.7 use dirservers before it's pulled down
  54. the directory?
  55. - put expiry date on onion-key, so people don't keep trying
  56. old ones that they could know are expired?
  57. * Leave on todo list, see if pre3 onion fixes helped enough.
  58. - should the running-routers list put unverified routers at the
  59. end?
  60. * Cosmetic, don't do it yet.
  61. - make advertised_server_mode() ORs fetch dirs more often.
  62. * not necessary yet.
  63. - Add a notion of nickname->Pubkey binding that's not 'verification'
  64. * eventually, only when needed
  65. - ORs use uniquer default nicknames
  66. * Don't worry about this for now
  67. - Handle full buffers without totally borking
  68. * do this eventually, no rush.
  69. more features, easy:
  70. o check the date in the http headers, compare for clock skew.
  71. o requiredentrynode vs preferredentrynode
  72. - per-month byte allowances
  73. * nick will spec something.
  74. - have a pool of circuits available, cannibalize them
  75. for your purposes (e.g. rendezvous, etc).
  76. * hold off on that.
  77. - node 'groups' that are known to be in the same zone of control
  78. * nick and roger will talk about it
  79. - do resolve before trying to attach the stream
  80. * don't do this for now.
  81. - if destination IP is running a tor node, extend a circuit there
  82. before sending begin.
  83. * don't do this for now. figure out how enclaves work. but do enclaves soon.
  84. more features, complex:
  85. - compress the directory. client sends http header
  86. "accept-transfer-encoding: gzip", server might send http header
  87. "transfer-encoding: gzip". ta-da.
  88. - grow a zlib dependency. keep a cached compressed directory.
  89. * nick will look into this. not critical priority.
  90. - Switch dirservers entries to config lines:
  91. - read in and parse each TrustedDir config line.
  92. - stop reading dirservers file.
  93. - add some default TrustedDir lines if none defined, or if
  94. no torrc.
  95. - remove notion of ->is_trusted_dir from the routerlist. that's
  96. no longer where you look.
  97. - clean up router parsing flow, since it's simpler now?
  98. - when checking signature on a directory, look it up in
  99. options.TrustedDirs, and make sure there's a descriptor
  100. with that nickname, whose key hashes to the fingerprint,
  101. and who correctly signed the directory.
  102. * nick will do the above
  103. - when fetching a directory, if you want a trusted one,
  104. choose from the trusteddir list.
  105. - which means keeping track of which ones are "up"
  106. - if you don't need a trusted one, choose from the routerinfo
  107. list if you have one, else from the trusteddir list.
  108. * roger will do the above
  109. - add a listener for a ui
  110. * nick chats with weasel
  111. - and a basic gui
  112. - Have clients and dirservers preserve reputation info over
  113. reboots.
  114. * continue not doing until we have something we need to preserve
  115. - users can set their bandwidth, or we auto-detect it:
  116. - advertised bandwidth defaults to 10KB
  117. o advertised bandwidth is the min of max seen in each direction
  118. in the past N seconds.
  119. o calculate this
  120. o not counting "local" connections
  121. - round detected bandwidth up to nearest 10KB
  122. - client software not upload descriptor until:
  123. - you've been running for an hour
  124. - it's sufficiently satisfied with its bandwidth
  125. - it decides it is reachable
  126. - start counting again if your IP ever changes.
  127. - never regenerate identity keys, for now.
  128. - you can set a bit for not-being-an-OR.
  129. * no need to do this yet. few people define their ORPort.
  130. - authdirserver lists you as running iff:
  131. - he can connect to you
  132. - he has successfully extended to you
  133. - you have sufficient mean-time-between-failures
  134. * keep doing nothing for now.
  135. blue sky:
  136. - Possible to get autoconf to easily install things into ~/.tor?
  137. ongoing:
  138. . rename/rearrange functions for what file they're in
  139. - generalize our transport: add transport.c in preparation for
  140. http, airhook, etc transport.
  141. NICK - investigate sctp for alternate transport.
  142. For September:
  143. NICK . Windows port
  144. o works as client
  145. - deal with pollhup / reached_eof on all platforms
  146. . robust as a client
  147. . works as server
  148. - can be configured
  149. - robust as a server
  150. . Usable as NT service
  151. - docs for building in win
  152. - installer
  153. - Docs
  154. . FAQ
  155. o overview of tor. how does it work, what's it do, pros and
  156. cons of using it, why should I use it, etc.
  157. - a howto tutorial with examples
  158. * put a stub on the wiki
  159. o tutorial: how to set up your own tor network
  160. - (need to not hardcode dirservers file in config.c)
  161. * this will be solved when we put dirservers in config lines
  162. - port forwarding howto for ipchains, etc
  163. * roger add to wiki of requests
  164. . correct, update, polish spec
  165. - document the exposed function api?
  166. o document what we mean by socks.
  167. NICK . packages
  168. . rpm
  169. * nick will look at the spec file
  170. - find a long-term rpm maintainer
  171. * roger will start guilting people
  172. - code
  173. - better warn/info messages
  174. o let tor do resolves.
  175. o extend socks4 to do resolves?
  176. o make script to ask tor for resolves
  177. - write howto for setting up tsocks, socat.
  178. - including on osx and win32
  179. - freecap handling
  180. - tsocks
  181. o gather patches, submit to maintainer
  182. * send him a reminder mail and see what's up.
  183. - intercept gethostbyname and others
  184. * add this to tsocks
  185. o do resolve via tor
  186. - redesign and thorough code revamp, with particular eye toward:
  187. - support half-open tcp connections
  188. - conn key rotation
  189. - other transports -- http, airhook
  190. - modular introduction mechanism
  191. - allow non-clique topology
  192. Other details and small and hard things:
  193. - tor should be able to have a pool of outgoing IP addresses
  194. that it is able to rotate through. (maybe)
  195. - tie into squid
  196. - hidserv offerers shouldn't need to define a SocksPort
  197. * figure out what breaks for this, and do it.
  198. - when the client fails to pick an intro point for a hidserv,
  199. it should refetch the hidserv desc.
  200. . should maybe make clients exit(1) when bad things happen?
  201. e.g. clock skew.
  202. - should retry exitpolicy end streams even if the end cell didn't
  203. resolve the address for you
  204. . Make logs handle it better when writing to them fails.
  205. o Dirserver shouldn't put you in running-routers list if you haven't
  206. uploaded a descriptor recently
  207. . Refactor: add own routerinfo to routerlist. Right now, only
  208. router_get_by_nickname knows about 'this router', as a hack to
  209. get circuit_launch_new to do the right thing.
  210. . Scrubbing proxies
  211. - Find an smtp proxy?
  212. . Get socks4a support into Mozilla
  213. - Need a relay teardown cell, separate from one-way ends.
  214. - Make it harder to circumvent bandwidth caps: look at number of bytes
  215. sent across sockets, not number sent inside TLS stream.
  216. - fix router_get_by_* functions so they can get ourselves too,
  217. and audit everything to make sure rend and intro points are
  218. just as likely to be us as not.
  219. ***************************Future tasks:****************************
  220. Rendezvous and hidden services:
  221. make it fast:
  222. - preemptively build and start rendezvous circs.
  223. - preemptively build n-1 hops of intro circs?
  224. - cannibalize general circs?
  225. make it reliable:
  226. - standby/hotswap/redundant services.
  227. - store stuff to disk? dirservers forget service descriptors when
  228. they restart; nodes offering hidden services forget their chosen
  229. intro points when they restart.
  230. make it robust:
  231. - auth mechanisms to let midpoint and bob selectively choose
  232. connection requests.
  233. make it scalable:
  234. - right now the hidserv store/lookup system is run by the dirservers;
  235. this won't scale.
  236. Tor scalability:
  237. Relax clique assumptions.
  238. Redesign how directories are handled.
  239. - Separate running-routers lookup from descriptor list lookup.
  240. - Resolve directory agreement somehow.
  241. - Cache directory on all servers.
  242. Find and remove bottlenecks
  243. - Address linear searches on e.g. circuit and connection lists.
  244. Reputation/memory system, so dirservers can measure people,
  245. and so other people can verify their measurements.
  246. - Need to measure via relay, so it's not distinguishable.
  247. Bandwidth-aware path selection. So people with T3's are picked
  248. more often than people with DSL.
  249. Reliability-aware node selection. So people who are stable are
  250. preferred for long-term circuits such as intro and rend circs,
  251. and general circs for irc, aim, ssh, etc.
  252. Let dissidents get to Tor servers via Tor users. ("Backbone model")
  253. Anonymity improvements:
  254. Is abandoning the circuit the only option when an extend fails, or
  255. can we do something without impacting anonymity too much?
  256. Is exiting from the middle of the circuit always a bad idea?
  257. Helper nodes. Decide how to use them to improve safety.
  258. DNS resolution: need to make tor support resolve requests. Need to write
  259. a script and an interface (including an extension to the socks
  260. protocol) so we can ask it to do resolve requests. Need to patch
  261. tsocks to intercept gethostbyname, else we'll continue leaking it.
  262. Improve path selection algorithms based on routing-zones paper. Be sure
  263. to start and end circuits in different ASs. Ideally, consider AS of
  264. source and destination -- maybe even enter and exit via nearby AS.
  265. Intermediate model, with some delays and mixing.
  266. Add defensive dropping regime?
  267. Make it more correct:
  268. Handle half-open connections: right now we don't support all TCP
  269. streams, at least according to the protocol. But we handle all that
  270. we've seen in the wild.
  271. Support IPv6.
  272. Efficiency/speed/robustness:
  273. Congestion control. Is our current design sufficient once we have heavy
  274. use? Need to measure and tweak, or maybe overhaul.
  275. Allow small cells and large cells on the same network?
  276. Cell buffering and resending. This will allow us to handle broken
  277. circuits as long as the endpoints don't break, plus will allow
  278. connection (tls session key) rotation.
  279. Implement Morphmix, so we can compare its behavior, complexity, etc.
  280. Use cpuworker for more heavy lifting.
  281. - Signing (and verifying) hidserv descriptors
  282. - Signing (and verifying) intro/rend requests
  283. - Signing (and verifying) router descriptors
  284. - Signing (and verifying) directories
  285. - Doing TLS handshake (this is very hard to separate out, though)
  286. Buffer size pool: allocate a maximum size for all buffers, not
  287. a maximum size for each buffer. So we don't have to give up as
  288. quickly (and kill the thickpipe!) when there's congestion.
  289. Exit node caching: tie into squid or other caching web proxy.
  290. Other transport. HTTP, udp, rdp, airhook, etc. May have to do our own
  291. link crypto, unless we can bully openssl into it.
  292. P2P Tor:
  293. Do all the scalability stuff above, first.
  294. Incentives to relay. Not so hard.
  295. Incentives to allow exit. Possibly quite hard.
  296. Sybil defenses without having a human bottleneck.
  297. How to gather random sample of nodes.
  298. How to handle nodelist recommendations.
  299. Consider incremental switches: a p2p tor with only 50 users has
  300. different anonymity properties than one with 10k users, and should
  301. be treated differently.