TODO 11 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248
  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 scalability:
  13. - Slightly smarter bandwidth management: use link capacity
  14. intelligently.
  15. - Handle full buffers without totally borking
  16. For dtor:
  17. NICK pre1:
  18. - make all ORs serve the directory too.
  19. - "AuthoritativeDir 1" for dirservers
  20. - non-authorative servers with dirport publish opt dircacheport
  21. - make clients read that and use it.
  22. - make clients able to read a normal dirport from non-trusted OR too
  23. - make ORs parse-and-keep the directory they pull down
  24. - authoritativedirservers should pull down directories from
  25. other authdirservers, to merge descriptors.
  26. - Have clients and dirservers preserve reputation info over
  27. reboots.
  28. - allow dirservers to serve running-router list separately.
  29. o "get /running-routers" will fetch just this.
  30. - actually make the clients use this sometimes.
  31. - distinguish directory-is-dirty from runninglist-is-dirty
  32. - ORs keep this too, and serve it
  33. - tor remembers descriptor-lists across reboots.
  34. - Packages define datadir as /var/lib/tor/. If no datadir is defined,
  35. then choose, make, and secure ~/.tor as datadir.
  36. - Contact info, pgp fingerprint, comments in router desc.
  37. - Add a ContactInfo line to torrc, which gets published in
  38. descriptor (as opt)
  39. - write tor version at the top of each log file
  40. pre2:
  41. - refer to things by key:
  42. - extend cells need ip:port:identitykeyhash.
  43. - also use this in intro points and rendezvous points, and
  44. hidserv descs.
  45. - figure out what to do about ip:port:differentkey
  46. ARMA - ORs connect on demand. attach circuits to new connections, keep
  47. create cells around somewhere, send destroy if fail.
  48. - nickname defaults to first piece of hostname
  49. - running-routers list refers to nickname if verified, else
  50. hash-base64'ed.
  51. pre3:
  52. - users can set their bandwidth, or we auto-detect it:
  53. - advertised bandwidth defaults to 10KB
  54. - advertised bandwidth is the min of max seen in each direction
  55. in the past N seconds.
  56. - not counting "local" connections
  57. - round detected bandwidth up to nearest 10KB
  58. - client software not upload descriptor until:
  59. - you've been running for an hour
  60. - it's sufficiently satisfied with its bandwidth
  61. - it decides it is reachable
  62. - start counting again if your IP ever changes.
  63. - never regenerate identity keys, for now.
  64. - you can set a bit for not-being-an-OR.
  65. - clients choose nodes proportional to advertised bandwidth
  66. - authdirserver includes descriptor and lists as running iff:
  67. - he can connect to you
  68. - he has successfully extended to you
  69. - he has sufficient mean-time-between-failures
  70. - add new "Middleman 1" config variable?
  71. - if torrc not found, exitpolicy reject *:*
  72. ongoing:
  73. . rename/rearrange functions for what file they're in
  74. - generalize our transport: add transport.c in preparation for
  75. http, airhook, etc transport.
  76. For September:
  77. NICK . Windows port
  78. o works as client
  79. - deal with pollhup / reached_eof on all platforms
  80. . robust as a client
  81. . works as server
  82. - can be configured
  83. - robust as a server
  84. . Usable as NT service
  85. - docs for building in win
  86. - installer
  87. - Docs
  88. - FAQ
  89. o overview of tor. how does it work, what's it do, pros and
  90. cons of using it, why should I use it, etc.
  91. - a howto tutorial with examples
  92. o tutorial: how to set up your own tor network
  93. - (need to not hardcode dirservers file in config.c)
  94. . correct, update, polish spec
  95. - document the exposed function api?
  96. - document what we mean by socks.
  97. NICK . packages
  98. . rpm
  99. - find a long-term rpm maintainer
  100. - code
  101. - better warn/info messages
  102. o let tor do resolves.
  103. o extend socks4 to do resolves?
  104. o make script to ask tor for resolves
  105. - tsocks
  106. - gather patches, submit to maintainer
  107. - intercept gethostbyname and others, do resolve via tor
  108. - redesign and thorough code revamp, with particular eye toward:
  109. - support half-open tcp connections
  110. - conn key rotation
  111. - other transports -- http, airhook
  112. - modular introduction mechanism
  113. - allow non-clique topology
  114. Other details and small and hard things:
  115. - tor should be able to have a pool of outgoing IP addresses
  116. that it is able to rotate through. (maybe)
  117. - tie into squid
  118. - buffer size pool, to let a few buffers grow huge or many buffers
  119. grow a bit
  120. - hidserv offerers shouldn't need to define a SocksPort
  121. - when the client fails to pick an intro point for a hidserv,
  122. it should refetch the hidserv desc.
  123. . should maybe make clients exit(1) when bad things happen?
  124. e.g. clock skew.
  125. - should retry exitpolicy end streams even if the end cell didn't
  126. resolve the address for you
  127. - Add '[...truncated]' or similar to truncated log entries (like the directory
  128. in connection_dir_process_inbuf()).
  129. . Make logs handle it better when writing to them fails.
  130. o Dirserver shouldn't put you in running-routers list if you haven't
  131. uploaded a descriptor recently
  132. . Refactor: add own routerinfo to routerlist. Right now, only
  133. router_get_by_nickname knows about 'this router', as a hack to
  134. get circuit_launch_new to do the right thing.
  135. . Scrubbing proxies
  136. - Find an smtp proxy?
  137. . Get socks4a support into Mozilla
  138. - Extend by hostname, not by IP.
  139. - Need a relay teardown cell, separate from one-way ends.
  140. - Make it harder to circumvent bandwidth caps: look at number of bytes
  141. sent across sockets, not number sent inside TLS stream.
  142. - fix router_get_by_* functions so they can get ourselves too,
  143. and audit everything to make sure rend and intro points are
  144. just as likely to be us as not.
  145. ***************************Future tasks:****************************
  146. Rendezvous and hidden services:
  147. make it fast:
  148. - preemptively build and start rendezvous circs.
  149. - preemptively build n-1 hops of intro circs?
  150. - cannibalize general circs?
  151. make it reliable:
  152. - standby/hotswap/redundant services.
  153. - store stuff to disk? dirservers forget service descriptors when
  154. they restart; nodes offering hidden services forget their chosen
  155. intro points when they restart.
  156. make it robust:
  157. - auth mechanisms to let midpoint and bob selectively choose
  158. connection requests.
  159. make it scalable:
  160. - right now the hidserv store/lookup system is run by the dirservers;
  161. this won't scale.
  162. Tor scalability:
  163. Relax clique assumptions.
  164. Redesign how directories are handled.
  165. - Separate running-routers lookup from descriptor list lookup.
  166. - Resolve directory agreement somehow.
  167. - Cache directory on all servers.
  168. Find and remove bottlenecks
  169. - Address linear searches on e.g. circuit and connection lists.
  170. Reputation/memory system, so dirservers can measure people,
  171. and so other people can verify their measurements.
  172. - Need to measure via relay, so it's not distinguishable.
  173. Bandwidth-aware path selection. So people with T3's are picked
  174. more often than people with DSL.
  175. Reliability-aware node selection. So people who are stable are
  176. preferred for long-term circuits such as intro and rend circs,
  177. and general circs for irc, aim, ssh, etc.
  178. Let dissidents get to Tor servers via Tor users. ("Backbone model")
  179. Anonymity improvements:
  180. Is abandoning the circuit the only option when an extend fails, or
  181. can we do something without impacting anonymity too much?
  182. Is exiting from the middle of the circuit always a bad idea?
  183. Helper nodes. Decide how to use them to improve safety.
  184. DNS resolution: need to make tor support resolve requests. Need to write
  185. a script and an interface (including an extension to the socks
  186. protocol) so we can ask it to do resolve requests. Need to patch
  187. tsocks to intercept gethostbyname, else we'll continue leaking it.
  188. Improve path selection algorithms based on routing-zones paper. Be sure
  189. to start and end circuits in different ASs. Ideally, consider AS of
  190. source and destination -- maybe even enter and exit via nearby AS.
  191. Intermediate model, with some delays and mixing.
  192. Add defensive dropping regime?
  193. Make it more correct:
  194. Handle half-open connections: right now we don't support all TCP
  195. streams, at least according to the protocol. But we handle all that
  196. we've seen in the wild.
  197. Support IPv6.
  198. Efficiency/speed/robustness:
  199. Congestion control. Is our current design sufficient once we have heavy
  200. use? Need to measure and tweak, or maybe overhaul.
  201. Allow small cells and large cells on the same network?
  202. Cell buffering and resending. This will allow us to handle broken
  203. circuits as long as the endpoints don't break, plus will allow
  204. connection (tls session key) rotation.
  205. Implement Morphmix, so we can compare its behavior, complexity, etc.
  206. Use cpuworker for more heavy lifting.
  207. - Signing (and verifying) hidserv descriptors
  208. - Signing (and verifying) intro/rend requests
  209. - Signing (and verifying) router descriptors
  210. - Signing (and verifying) directories
  211. - Doing TLS handshake (this is very hard to separate out, though)
  212. Buffer size pool: allocate a maximum size for all buffers, not
  213. a maximum size for each buffer. So we don't have to give up as
  214. quickly (and kill the thickpipe!) when there's congestion.
  215. Exit node caching: tie into squid or other caching web proxy.
  216. Other transport. HTTP, udp, rdp, airhook, etc. May have to do our own
  217. link crypto, unless we can bully openssl into it.
  218. P2P Tor:
  219. Do all the scalability stuff above, first.
  220. Incentives to relay. Not so hard.
  221. Incentives to allow exit. Possibly quite hard.
  222. Sybil defenses without having a human bottleneck.
  223. How to gather random sample of nodes.
  224. How to handle nodelist recommendations.
  225. Consider incremental switches: a p2p tor with only 50 users has
  226. different anonymity properties than one with 10k users, and should
  227. be treated differently.