TODO 12 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274
  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. misc:
  13. . refer to things by key:
  14. - also use this in intro points and rendezvous points, and
  15. hidserv descs. [XXXX This isn't enough.]
  16. - figure out what to do about ip:port:differentkey
  17. - users can set their bandwidth, or we auto-detect it:
  18. - advertised bandwidth defaults to 10KB
  19. o advertised bandwidth is the min of max seen in each direction
  20. in the past N seconds.
  21. o calculate this
  22. o not counting "local" connections
  23. - round detected bandwidth up to nearest 10KB
  24. - client software not upload descriptor until:
  25. - you've been running for an hour
  26. - it's sufficiently satisfied with its bandwidth
  27. - it decides it is reachable
  28. - start counting again if your IP ever changes.
  29. - never regenerate identity keys, for now.
  30. - you can set a bit for not-being-an-OR.
  31. NICK - Reputation info needs to give better weight to recent events than
  32. very old ones.
  33. - Have clients and dirservers preserve reputation info over
  34. reboots.
  35. - clients choose nodes proportional to advertised bandwidth
  36. - and lists as running iff:
  37. - he can connect to you
  38. - he has successfully extended to you
  39. - you have sufficient mean-time-between-failures
  40. docs:
  41. - faq and doc-wiki
  42. - knoppix distro
  43. - win32 installer using privoxy's installer
  44. bug fixes, necessary:
  45. - Why is the first entry of kill -USR1 a router with a 0 key?
  46. - why does common/util.c build-depend on or/or.h ?
  47. bug fixes, might be handy:
  48. - put expiry date on onion-key, so people don't keep trying
  49. old ones that they could know are expired?
  50. - should the running-routers list put unverified routers at the
  51. end?
  52. - make advertised_server_mode() ORs fetch dirs more often.
  53. - if a begin failed due to exit policy, but we believe the IP
  54. should have been allowed, switch that router to exitpolicy
  55. reject *:* until we get our next directory.
  56. - Tors deal appropriately when a newly-verified router has the
  57. same nickname as another router they know about
  58. - ORs use uniquer default nicknames
  59. - Handle full buffers without totally borking
  60. - Add '[...truncated]' or similar to truncated log entries (like the directory
  61. in connection_dir_process_inbuf()).
  62. more features, easy:
  63. - check the date in the http headers, compare for clock skew.
  64. - requiredentrynode vs preferredentrynode
  65. - per-month byte allowances
  66. - tor-resolve needs a man page
  67. - "AllowUnverifiedRouters" config option
  68. - Parse it into a bitvector
  69. - Consider it when picking nodes for your path
  70. - have a pool of circuits available, cannibalize them
  71. for your purposes (e.g. rendezvous, etc).
  72. more features, complex:
  73. - defining an ORPort isn't necessary anymore, if you use
  74. ORAddress:port. Same with DirPort, SocksPort.
  75. - compress the directory. client sends http header
  76. "accept-transfer-encoding: gzip", server might send http header
  77. "transfer-encoding: gzip". ta-da.
  78. - grow a zlib dependency. keep a cached compressed directory.
  79. - Switch dirservers entries to config lines:
  80. - read in and parse each TrustedDir config line.
  81. - stop reading dirservers file.
  82. - add some default TrustedDir lines if none defined, or if
  83. no torrc.
  84. - remove notion of ->is_trusted_dir from the routerlist. that's
  85. no longer where you look.
  86. - clean up router parsing flow, since it's simpler now?
  87. - when checking signature on a directory, look it up in
  88. options.TrustedDirs, and make sure there's a descriptor
  89. with that nickname, whose key hashes to the fingerprint,
  90. and who correctly signed the directory.
  91. - when fetching a directory, if you want a trusted one,
  92. choose from the trusteddir list.
  93. - which means keeping track of which ones are "up"
  94. - if you don't need a trusted one, choose from the routerinfo
  95. list if you have one, else from the trusteddir list.
  96. - add a listener for a ui
  97. - and a basic gui
  98. blue sky:
  99. - Possible to get autoconf to easily install things into ~/.tor?
  100. ongoing:
  101. . rename/rearrange functions for what file they're in
  102. - generalize our transport: add transport.c in preparation for
  103. http, airhook, etc transport.
  104. For September:
  105. NICK . Windows port
  106. o works as client
  107. - deal with pollhup / reached_eof on all platforms
  108. . robust as a client
  109. . works as server
  110. - can be configured
  111. - robust as a server
  112. . Usable as NT service
  113. - docs for building in win
  114. - installer
  115. - Docs
  116. - FAQ
  117. o overview of tor. how does it work, what's it do, pros and
  118. cons of using it, why should I use it, etc.
  119. - a howto tutorial with examples
  120. o tutorial: how to set up your own tor network
  121. - (need to not hardcode dirservers file in config.c)
  122. . correct, update, polish spec
  123. - document the exposed function api?
  124. o document what we mean by socks.
  125. NICK . packages
  126. . rpm
  127. - find a long-term rpm maintainer
  128. - code
  129. - better warn/info messages
  130. o let tor do resolves.
  131. o extend socks4 to do resolves?
  132. o make script to ask tor for resolves
  133. - tsocks
  134. o gather patches, submit to maintainer
  135. - intercept gethostbyname and others
  136. o do resolve via tor
  137. - redesign and thorough code revamp, with particular eye toward:
  138. - support half-open tcp connections
  139. - conn key rotation
  140. - other transports -- http, airhook
  141. - modular introduction mechanism
  142. - allow non-clique topology
  143. Other details and small and hard things:
  144. - tor should be able to have a pool of outgoing IP addresses
  145. that it is able to rotate through. (maybe)
  146. - tie into squid
  147. - hidserv offerers shouldn't need to define a SocksPort
  148. - when the client fails to pick an intro point for a hidserv,
  149. it should refetch the hidserv desc.
  150. . should maybe make clients exit(1) when bad things happen?
  151. e.g. clock skew.
  152. - should retry exitpolicy end streams even if the end cell didn't
  153. resolve the address for you
  154. . Make logs handle it better when writing to them fails.
  155. o Dirserver shouldn't put you in running-routers list if you haven't
  156. uploaded a descriptor recently
  157. . Refactor: add own routerinfo to routerlist. Right now, only
  158. router_get_by_nickname knows about 'this router', as a hack to
  159. get circuit_launch_new to do the right thing.
  160. . Scrubbing proxies
  161. - Find an smtp proxy?
  162. . Get socks4a support into Mozilla
  163. - Need a relay teardown cell, separate from one-way ends.
  164. - Make it harder to circumvent bandwidth caps: look at number of bytes
  165. sent across sockets, not number sent inside TLS stream.
  166. - fix router_get_by_* functions so they can get ourselves too,
  167. and audit everything to make sure rend and intro points are
  168. just as likely to be us as not.
  169. ***************************Future tasks:****************************
  170. Rendezvous and hidden services:
  171. make it fast:
  172. - preemptively build and start rendezvous circs.
  173. - preemptively build n-1 hops of intro circs?
  174. - cannibalize general circs?
  175. make it reliable:
  176. - standby/hotswap/redundant services.
  177. - store stuff to disk? dirservers forget service descriptors when
  178. they restart; nodes offering hidden services forget their chosen
  179. intro points when they restart.
  180. make it robust:
  181. - auth mechanisms to let midpoint and bob selectively choose
  182. connection requests.
  183. make it scalable:
  184. - right now the hidserv store/lookup system is run by the dirservers;
  185. this won't scale.
  186. Tor scalability:
  187. Relax clique assumptions.
  188. Redesign how directories are handled.
  189. - Separate running-routers lookup from descriptor list lookup.
  190. - Resolve directory agreement somehow.
  191. - Cache directory on all servers.
  192. Find and remove bottlenecks
  193. - Address linear searches on e.g. circuit and connection lists.
  194. Reputation/memory system, so dirservers can measure people,
  195. and so other people can verify their measurements.
  196. - Need to measure via relay, so it's not distinguishable.
  197. Bandwidth-aware path selection. So people with T3's are picked
  198. more often than people with DSL.
  199. Reliability-aware node selection. So people who are stable are
  200. preferred for long-term circuits such as intro and rend circs,
  201. and general circs for irc, aim, ssh, etc.
  202. Let dissidents get to Tor servers via Tor users. ("Backbone model")
  203. Anonymity improvements:
  204. Is abandoning the circuit the only option when an extend fails, or
  205. can we do something without impacting anonymity too much?
  206. Is exiting from the middle of the circuit always a bad idea?
  207. Helper nodes. Decide how to use them to improve safety.
  208. DNS resolution: need to make tor support resolve requests. Need to write
  209. a script and an interface (including an extension to the socks
  210. protocol) so we can ask it to do resolve requests. Need to patch
  211. tsocks to intercept gethostbyname, else we'll continue leaking it.
  212. Improve path selection algorithms based on routing-zones paper. Be sure
  213. to start and end circuits in different ASs. Ideally, consider AS of
  214. source and destination -- maybe even enter and exit via nearby AS.
  215. Intermediate model, with some delays and mixing.
  216. Add defensive dropping regime?
  217. Make it more correct:
  218. Handle half-open connections: right now we don't support all TCP
  219. streams, at least according to the protocol. But we handle all that
  220. we've seen in the wild.
  221. Support IPv6.
  222. Efficiency/speed/robustness:
  223. Congestion control. Is our current design sufficient once we have heavy
  224. use? Need to measure and tweak, or maybe overhaul.
  225. Allow small cells and large cells on the same network?
  226. Cell buffering and resending. This will allow us to handle broken
  227. circuits as long as the endpoints don't break, plus will allow
  228. connection (tls session key) rotation.
  229. Implement Morphmix, so we can compare its behavior, complexity, etc.
  230. Use cpuworker for more heavy lifting.
  231. - Signing (and verifying) hidserv descriptors
  232. - Signing (and verifying) intro/rend requests
  233. - Signing (and verifying) router descriptors
  234. - Signing (and verifying) directories
  235. - Doing TLS handshake (this is very hard to separate out, though)
  236. Buffer size pool: allocate a maximum size for all buffers, not
  237. a maximum size for each buffer. So we don't have to give up as
  238. quickly (and kill the thickpipe!) when there's congestion.
  239. Exit node caching: tie into squid or other caching web proxy.
  240. Other transport. HTTP, udp, rdp, airhook, etc. May have to do our own
  241. link crypto, unless we can bully openssl into it.
  242. P2P Tor:
  243. Do all the scalability stuff above, first.
  244. Incentives to relay. Not so hard.
  245. Incentives to allow exit. Possibly quite hard.
  246. Sybil defenses without having a human bottleneck.
  247. How to gather random sample of nodes.
  248. How to handle nodelist recommendations.
  249. Consider incremental switches: a p2p tor with only 50 users has
  250. different anonymity properties than one with 10k users, and should
  251. be treated differently.