TODO 12 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292
  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. For 0.0.9:
  13. N&R- bring tor-spec up to date
  14. N . cache and serve running-routers on other nodes?
  15. . cache running-routers
  16. - download running-routers from servers running rc5-cvs or later
  17. N - pump up periods for fetching things; figure out how to do this
  18. backward-compatibily, so that people who did set dirfetchpostperiod
  19. get the right behavior.
  20. N - Adapt version parsing code to handle new version scheme; document new
  21. version scheme.
  22. N&R. make loglevels info,debug less noisy
  23. R - fix dfc/weasel's intro point bug
  24. R - add goodell's .exit tld
  25. Beyond 0.0.9:
  26. - server descriptor declares min log level, clients avoid servers
  27. that are too loggy.
  28. N - Clean up NT service code
  29. N - OS X package (and bundle?)
  30. - controller should have 'getinfo' command to query about rephist,
  31. about rendezvous status, etc.
  32. - allow transition from ORPort to !ORPort, and back
  33. R . bandwidth buckets for write as well as read.
  34. - Limit to 2 dir, 2 OR, N SOCKS connections per IP.
  35. - Implement If-Modified-Since for directories.
  36. - Make more configuration variables into CSVs.
  37. N - Handle rendezvousing with unverified nodes.
  38. - Specify: Stick rendezvous point's key in INTRODUCE cell.
  39. Bob should _always_ use key from INTRODUCE cell.
  40. - Implement.
  41. R - figure out enclaves, e.g. so we know what to recommend that people
  42. do, and so running a tor server on your website is helpful.
  43. - Do enclaves for same IP only.
  44. - Resolve first, then if IP is an OR, connect to next guy.
  45. N . the user interface interface
  46. - Implement a trivial fun gui.
  47. N - add ipv6 support.
  48. - Spec issue: if a resolve returns an IP4 and an IP6 address,
  49. which to use?
  50. N&R - Update Spec
  51. R X learn from ben about his openssl-reinitialization-trick to
  52. rotate tls keys without making new connections.
  53. - Do something to prevent spurious EXTEND cells from making middleman
  54. nodes connect all over. Rate-limit failed connections, perhaps?
  55. - christian grothoff's attack of infinite-length circuit.
  56. the solution is to have a separate 'extend-data' cell type
  57. which is used for the first N data cells, and only
  58. extend-data cells can be extend requests.
  59. - have a pool of circuits available, cannibalize them
  60. for your purposes (e.g. rendezvous, etc).
  61. - Once we have a trusted directory on port 80, stop falling back to
  62. forbidden ports when fascistfirewall blocks all good dirservers.
  63. o fix sprintf's to snprintf's?
  64. . Make intro points and rendezvous points accept $KEYID in addition
  65. to nicknames.
  66. o Specify
  67. o Implement parsing
  68. - Generate new formats (Not till 007 is dead)
  69. - Facility to automatically choose long-term helper nodes; perhaps
  70. on by default for hidden services.
  71. o Make command-line strict about checking options; make only certain
  72. option prefixes work.
  73. - Rate-limit OR and directory connections overall and per-IP and
  74. maybe per subnet.
  75. D put expiry date on onion-key, so people don't keep trying
  76. old ones that they could know are expired?
  77. * Leave on todo list, see if pre3 onion fixes helped enough.
  78. D should the running-routers list put unverified routers at the
  79. end?
  80. * Cosmetic, don't do it yet.
  81. D make advertised_server_mode() ORs fetch dirs more often.
  82. * not necessary yet.
  83. D Add a notion of nickname->Pubkey binding that's not 'verification'
  84. * eventually, only when needed
  85. D ORs use uniquer default nicknames
  86. * Don't worry about this for now
  87. D Handle full buffers without totally borking
  88. * do this eventually, no rush.
  89. D if destination IP is running a tor node, extend a circuit there
  90. before sending begin.
  91. * don't do this for now. figure out how enclaves work. but do
  92. enclaves soon.
  93. - Support egd or other non-OS-integrated strong entropy sources
  94. more features, complex:
  95. - password protection for on-disk identity key
  96. - Have clients and dirservers preserve reputation info over
  97. reboots.
  98. * continue not doing until we have something we need to preserve
  99. - round detected bandwidth up to nearest 10KB?
  100. - client software not upload descriptor until:
  101. - you've been running for an hour
  102. - it's sufficiently satisfied with its bandwidth
  103. - it decides it is reachable
  104. - start counting again if your IP ever changes.
  105. - never regenerate identity keys, for now.
  106. - you can set a bit for not-being-an-OR.
  107. * no need to do this yet. few people define their ORPort.
  108. - authdirserver lists you as running iff:
  109. - he can connect to you
  110. - he has successfully extended to you
  111. - you have sufficient mean-time-between-failures
  112. * keep doing nothing for now.
  113. - Include HTTP status messages in logging (see parse_http_response).
  114. blue sky:
  115. - Possible to get autoconf to easily install things into ~/.tor?
  116. ongoing:
  117. . rename/rearrange functions for what file they're in
  118. - generalize our transport: add transport.c in preparation for
  119. http, airhook, etc transport.
  120. o investigate sctp for alternate transport.
  121. For September:
  122. N . Windows port
  123. o works as client
  124. - deal with pollhup / reached_eof on all platforms
  125. . robust as a client
  126. . works as server
  127. - can be configured
  128. - robust as a server
  129. . Usable as NT service
  130. - docs for building in win
  131. o installer, including all needed libs.
  132. - and including privoxy
  133. - and including a sockscap equivalent
  134. - Docs
  135. . FAQ
  136. - a howto tutorial with examples
  137. * put a stub on the wiki
  138. o tutorial: how to set up your own tor network
  139. o (need to not hardcode dirservers file in config.c)
  140. o Make tutorial reflect this.
  141. . port forwarding howto for ipchains, etc
  142. . correct, update, polish spec
  143. - document the exposed function api?
  144. - Document where we differ from tor-design
  145. . packages
  146. . find a long-term rpm maintainer
  147. - code
  148. - better warn/info messages
  149. - write howto for setting up tsocks, socat.
  150. - including on osx and win32
  151. - freecap handling
  152. - tsocks
  153. o gather patches, submit to maintainer
  154. * send him a reminder mail and see what's up.
  155. - intercept gethostbyname and others
  156. * add this to tsocks
  157. o do resolve via tor
  158. - redesign and thorough code revamp, with particular eye toward:
  159. - support half-open tcp connections
  160. - conn key rotation
  161. - other transports -- http, airhook
  162. - modular introduction mechanism
  163. - allow non-clique topology
  164. Other details and small and hard things:
  165. - tor should be able to have a pool of outgoing IP addresses
  166. that it is able to rotate through. (maybe)
  167. - tie into squid
  168. - hidserv offerers shouldn't need to define a SocksPort
  169. * figure out what breaks for this, and do it.
  170. - when the client fails to pick an intro point for a hidserv,
  171. it should refetch the hidserv desc.
  172. . should maybe make clients exit(1) when bad things happen?
  173. e.g. clock skew.
  174. - should retry exitpolicy end streams even if the end cell didn't
  175. resolve the address for you
  176. o Make logs handle it better when writing to them fails.
  177. o Dirserver shouldn't put you in running-routers list if you haven't
  178. uploaded a descriptor recently
  179. . Refactor: add own routerinfo to routerlist. Right now, only
  180. router_get_by_nickname knows about 'this router', as a hack to
  181. get circuit_launch_new to do the right thing.
  182. . Scrubbing proxies
  183. - Find an smtp proxy?
  184. . Get socks4a support into Mozilla
  185. - Need a relay teardown cell, separate from one-way ends.
  186. - Make it harder to circumvent bandwidth caps: look at number of bytes
  187. sent across sockets, not number sent inside TLS stream.
  188. - fix router_get_by_* functions so they can get ourselves too,
  189. and audit everything to make sure rend and intro points are
  190. just as likely to be us as not.
  191. ***************************Future tasks:****************************
  192. Rendezvous and hidden services:
  193. make it fast:
  194. - preemptively build and start rendezvous circs.
  195. - preemptively build n-1 hops of intro circs?
  196. - cannibalize general circs?
  197. make it reliable:
  198. - standby/hotswap/redundant services.
  199. - store stuff to disk? dirservers forget service descriptors when
  200. they restart; nodes offering hidden services forget their chosen
  201. intro points when they restart.
  202. make it robust:
  203. - auth mechanisms to let midpoint and bob selectively choose
  204. connection requests.
  205. make it scalable:
  206. - right now the hidserv store/lookup system is run by the dirservers;
  207. this won't scale.
  208. Tor scalability:
  209. Relax clique assumptions.
  210. Redesign how directories are handled.
  211. o Separate running-routers lookup from descriptor list lookup.
  212. - Resolve directory agreement somehow.
  213. o Cache directory on all servers.
  214. Find and remove bottlenecks
  215. - Address linear searches on e.g. circuit and connection lists.
  216. Reputation/memory system, so dirservers can measure people,
  217. and so other people can verify their measurements.
  218. - Need to measure via relay, so it's not distinguishable.
  219. Bandwidth-aware path selection. So people with T3's are picked
  220. more often than people with DSL.
  221. Reliability-aware node selection. So people who are stable are
  222. preferred for long-term circuits such as intro and rend circs,
  223. and general circs for irc, aim, ssh, etc.
  224. Let dissidents get to Tor servers via Tor users. ("Backbone model")
  225. Anonymity improvements:
  226. Is abandoning the circuit the only option when an extend fails, or
  227. can we do something without impacting anonymity too much?
  228. Is exiting from the middle of the circuit always a bad idea?
  229. Helper nodes. Decide how to use them to improve safety.
  230. DNS resolution: need to make tor support resolve requests. Need to write
  231. a script and an interface (including an extension to the socks
  232. protocol) so we can ask it to do resolve requests. Need to patch
  233. tsocks to intercept gethostbyname, else we'll continue leaking it.
  234. Improve path selection algorithms based on routing-zones paper. Be sure
  235. to start and end circuits in different ASs. Ideally, consider AS of
  236. source and destination -- maybe even enter and exit via nearby AS.
  237. Intermediate model, with some delays and mixing.
  238. Add defensive dropping regime?
  239. Make it more correct:
  240. Handle half-open connections: right now we don't support all TCP
  241. streams, at least according to the protocol. But we handle all that
  242. we've seen in the wild.
  243. Support IPv6.
  244. Efficiency/speed/robustness:
  245. Congestion control. Is our current design sufficient once we have heavy
  246. use? Need to measure and tweak, or maybe overhaul.
  247. Allow small cells and large cells on the same network?
  248. Cell buffering and resending. This will allow us to handle broken
  249. circuits as long as the endpoints don't break, plus will allow
  250. connection (tls session key) rotation.
  251. Implement Morphmix, so we can compare its behavior, complexity, etc.
  252. Use cpuworker for more heavy lifting.
  253. - Signing (and verifying) hidserv descriptors
  254. - Signing (and verifying) intro/rend requests
  255. - Signing (and verifying) router descriptors
  256. - Signing (and verifying) directories
  257. - Doing TLS handshake (this is very hard to separate out, though)
  258. Buffer size pool: allocate a maximum size for all buffers, not
  259. a maximum size for each buffer. So we don't have to give up as
  260. quickly (and kill the thickpipe!) when there's congestion.
  261. Exit node caching: tie into squid or other caching web proxy.
  262. Other transport. HTTP, udp, rdp, airhook, etc. May have to do our own
  263. link crypto, unless we can bully openssl into it.
  264. P2P Tor:
  265. Do all the scalability stuff above, first.
  266. Incentives to relay. Not so hard.
  267. Incentives to allow exit. Possibly quite hard.
  268. Sybil defenses without having a human bottleneck.
  269. How to gather random sample of nodes.
  270. How to handle nodelist recommendations.
  271. Consider incremental switches: a p2p tor with only 50 users has
  272. different anonymity properties than one with 10k users, and should
  273. be treated differently.