TODO 15 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346
  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. 0.0.9:
  13. o Fix OutboundBindAddress
  14. o Config defaults should be consistant with config file and no
  15. config file.
  16. o write instructions for port-forwarding directives or programs
  17. to let people run on ports 80 and 443 without needing to bind
  18. tor to them.
  19. o clean up all the comma-separated stuff (eg exit policies) into
  20. smartlists.
  21. o investigate sctp for alternate transport.
  22. o Document all undocumented options, or mark them as undocumented
  23. in the source.
  24. o bandwidth buckets for write as well as read.
  25. . Cached-directory changes:
  26. o make clients store the cached-directory to disk,
  27. o and use it when they startup, so they don't need to bootstrap
  28. from the authdirservers every time they start.
  29. - also, once we've reduced authdirserver entries to config
  30. lines, we can have lines that list cacheddirservers too.
  31. . compress the directory.
  32. o Implement gzip/zlib wrappers
  33. o Compress directories as they're cached/generated
  34. o When requested, give a compressed directory.
  35. o Decompress incoming HTTP based on Content-Encoding
  36. - Once dirservers are running new code, make clients
  37. request compressed directories. (Alternative: Switch
  38. to HTTP/1.1 and use Allowed-Encoding. Is that really
  39. what we want?)
  40. N - switch dirservers entries to config lines.
  41. N - let tor clients use http proxies for dir fetching
  42. N - per-month byte allowances.
  43. Nr - figure out how to handle rendezvousing with unverified nodes.
  44. Nr - figure out enclaves, e.g. so we know what to recommend that people
  45. do, and so running a tor server on your website is helpful.
  46. - node 'groups' that are known to be in the same zone of control.
  47. - let tor servers use http proxies for port 80 exits
  48. - the user interface interface
  49. - add ipv6 support.
  50. - learn from ben about his openssl-reinitialization-trick to
  51. rotate tls keys without making new connections.
  52. D nt services on win32.
  53. 0.0.8:
  54. - fix sprintf's to snprintf's?
  55. o Make it work on win32 with no $home
  56. o Don't crash.
  57. o Put files someplace reasonable.
  58. o Why is the first entry of kill -USR1 a router with a 0 key?
  59. o Tors deal appropriately when a newly-verified router has the
  60. same nickname as another router they know about
  61. X put ip:port:keyhash in intro points, rendezvous points,
  62. and hidserv descriptors.
  63. . Make intro points and rendezvous points accept $KEYID in addition
  64. to nicknames.
  65. o Specify
  66. o Implement parsing
  67. - Generate new formats (Not till 007 is dead)
  68. NICK . unify similar config entries that need to be split. put them
  69. into a smartlist, and have things take a smartlist.
  70. - figure out what to do when somebody asks to extend to
  71. ip:port:differentkey
  72. * reject it. assuming this is as dumb as it sounds.
  73. - make loglevel info less noisy
  74. bug fixes, might be handy:
  75. - the directory servers complain a lot about people using the
  76. old key. does 0.0.7 use dirservers before it's pulled down
  77. the directory?
  78. - put expiry date on onion-key, so people don't keep trying
  79. old ones that they could know are expired?
  80. * Leave on todo list, see if pre3 onion fixes helped enough.
  81. - should the running-routers list put unverified routers at the
  82. end?
  83. * Cosmetic, don't do it yet.
  84. - make advertised_server_mode() ORs fetch dirs more often.
  85. * not necessary yet.
  86. - Add a notion of nickname->Pubkey binding that's not 'verification'
  87. * eventually, only when needed
  88. - ORs use uniquer default nicknames
  89. * Don't worry about this for now
  90. - Handle full buffers without totally borking
  91. * do this eventually, no rush.
  92. more features, easy:
  93. - per-month byte allowances
  94. * nick will spec something.
  95. - have a pool of circuits available, cannibalize them
  96. for your purposes (e.g. rendezvous, etc).
  97. * hold off on that.
  98. - node 'groups' that are known to be in the same zone of control
  99. * nick and roger will talk about it
  100. - do resolve before trying to attach the stream
  101. * don't do this for now.
  102. - if destination IP is running a tor node, extend a circuit there
  103. before sending begin.
  104. * don't do this for now. figure out how enclaves work. but do enclaves soon.
  105. - Track max ten-second b/w ever seen, to show operator
  106. more features, complex:
  107. - compress the directory. client sends http header
  108. "accept-transfer-encoding: gzip", server might send http header
  109. "transfer-encoding: gzip". ta-da.
  110. - grow a zlib dependency. keep a cached compressed directory.
  111. * nick will look into this. not critical priority.
  112. - Switch dirservers entries to config lines:
  113. - read in and parse each TrustedDir config line.
  114. - stop reading dirservers file.
  115. - add some default TrustedDir lines if none defined, or if
  116. no torrc.
  117. - remove notion of ->is_trusted_dir from the routerlist. that's
  118. no longer where you look.
  119. - clean up router parsing flow, since it's simpler now?
  120. - when checking signature on a directory, look it up in
  121. options.TrustedDirs, and make sure there's a descriptor
  122. with that nickname, whose key hashes to the fingerprint,
  123. and who correctly signed the directory.
  124. * nick will do the above
  125. - when fetching a directory, if you want a trusted one,
  126. choose from the trusteddir list.
  127. - which means keeping track of which ones are "up"
  128. - if you don't need a trusted one, choose from the routerinfo
  129. list if you have one, else from the trusteddir list.
  130. * roger will do the above
  131. - add a listener for a ui
  132. * nick chats with weasel
  133. - and a basic gui
  134. - Have clients and dirservers preserve reputation info over
  135. reboots.
  136. * continue not doing until we have something we need to preserve
  137. - users can set their bandwidth, or we auto-detect it:
  138. - advertised bandwidth defaults to 10KB
  139. o advertised bandwidth is the min of max seen in each direction
  140. in the past N seconds.
  141. o calculate this
  142. o not counting "local" connections
  143. - round detected bandwidth up to nearest 10KB
  144. - client software not upload descriptor until:
  145. - you've been running for an hour
  146. - it's sufficiently satisfied with its bandwidth
  147. - it decides it is reachable
  148. - start counting again if your IP ever changes.
  149. - never regenerate identity keys, for now.
  150. - you can set a bit for not-being-an-OR.
  151. * no need to do this yet. few people define their ORPort.
  152. - authdirserver lists you as running iff:
  153. - he can connect to you
  154. - he has successfully extended to you
  155. - you have sufficient mean-time-between-failures
  156. * keep doing nothing for now.
  157. blue sky:
  158. - Possible to get autoconf to easily install things into ~/.tor?
  159. ongoing:
  160. . rename/rearrange functions for what file they're in
  161. - generalize our transport: add transport.c in preparation for
  162. http, airhook, etc transport.
  163. NICK - investigate sctp for alternate transport.
  164. For September:
  165. NICK . Windows port
  166. o works as client
  167. - deal with pollhup / reached_eof on all platforms
  168. . robust as a client
  169. . works as server
  170. - can be configured
  171. - robust as a server
  172. . Usable as NT service
  173. - docs for building in win
  174. - installer
  175. - Docs
  176. . FAQ
  177. o overview of tor. how does it work, what's it do, pros and
  178. cons of using it, why should I use it, etc.
  179. - a howto tutorial with examples
  180. * put a stub on the wiki
  181. o tutorial: how to set up your own tor network
  182. - (need to not hardcode dirservers file in config.c)
  183. * this will be solved when we put dirservers in config lines
  184. - port forwarding howto for ipchains, etc
  185. * roger add to wiki of requests
  186. . correct, update, polish spec
  187. - document the exposed function api?
  188. o document what we mean by socks.
  189. NICK . packages
  190. . rpm
  191. * nick will look at the spec file
  192. - find a long-term rpm maintainer
  193. * roger will start guilting people
  194. - code
  195. - better warn/info messages
  196. o let tor do resolves.
  197. o extend socks4 to do resolves?
  198. o make script to ask tor for resolves
  199. - write howto for setting up tsocks, socat.
  200. - including on osx and win32
  201. - freecap handling
  202. - tsocks
  203. o gather patches, submit to maintainer
  204. * send him a reminder mail and see what's up.
  205. - intercept gethostbyname and others
  206. * add this to tsocks
  207. o do resolve via tor
  208. - redesign and thorough code revamp, with particular eye toward:
  209. - support half-open tcp connections
  210. - conn key rotation
  211. - other transports -- http, airhook
  212. - modular introduction mechanism
  213. - allow non-clique topology
  214. Other details and small and hard things:
  215. - tor should be able to have a pool of outgoing IP addresses
  216. that it is able to rotate through. (maybe)
  217. - tie into squid
  218. - hidserv offerers shouldn't need to define a SocksPort
  219. * figure out what breaks for this, and do it.
  220. - when the client fails to pick an intro point for a hidserv,
  221. it should refetch the hidserv desc.
  222. . should maybe make clients exit(1) when bad things happen?
  223. e.g. clock skew.
  224. - should retry exitpolicy end streams even if the end cell didn't
  225. resolve the address for you
  226. . Make logs handle it better when writing to them fails.
  227. o Dirserver shouldn't put you in running-routers list if you haven't
  228. uploaded a descriptor recently
  229. . Refactor: add own routerinfo to routerlist. Right now, only
  230. router_get_by_nickname knows about 'this router', as a hack to
  231. get circuit_launch_new to do the right thing.
  232. . Scrubbing proxies
  233. - Find an smtp proxy?
  234. . Get socks4a support into Mozilla
  235. - Need a relay teardown cell, separate from one-way ends.
  236. - Make it harder to circumvent bandwidth caps: look at number of bytes
  237. sent across sockets, not number sent inside TLS stream.
  238. - fix router_get_by_* functions so they can get ourselves too,
  239. and audit everything to make sure rend and intro points are
  240. just as likely to be us as not.
  241. ***************************Future tasks:****************************
  242. Rendezvous and hidden services:
  243. make it fast:
  244. - preemptively build and start rendezvous circs.
  245. - preemptively build n-1 hops of intro circs?
  246. - cannibalize general circs?
  247. make it reliable:
  248. - standby/hotswap/redundant services.
  249. - store stuff to disk? dirservers forget service descriptors when
  250. they restart; nodes offering hidden services forget their chosen
  251. intro points when they restart.
  252. make it robust:
  253. - auth mechanisms to let midpoint and bob selectively choose
  254. connection requests.
  255. make it scalable:
  256. - right now the hidserv store/lookup system is run by the dirservers;
  257. this won't scale.
  258. Tor scalability:
  259. Relax clique assumptions.
  260. Redesign how directories are handled.
  261. - Separate running-routers lookup from descriptor list lookup.
  262. - Resolve directory agreement somehow.
  263. - Cache directory on all servers.
  264. Find and remove bottlenecks
  265. - Address linear searches on e.g. circuit and connection lists.
  266. Reputation/memory system, so dirservers can measure people,
  267. and so other people can verify their measurements.
  268. - Need to measure via relay, so it's not distinguishable.
  269. Bandwidth-aware path selection. So people with T3's are picked
  270. more often than people with DSL.
  271. Reliability-aware node selection. So people who are stable are
  272. preferred for long-term circuits such as intro and rend circs,
  273. and general circs for irc, aim, ssh, etc.
  274. Let dissidents get to Tor servers via Tor users. ("Backbone model")
  275. Anonymity improvements:
  276. Is abandoning the circuit the only option when an extend fails, or
  277. can we do something without impacting anonymity too much?
  278. Is exiting from the middle of the circuit always a bad idea?
  279. Helper nodes. Decide how to use them to improve safety.
  280. DNS resolution: need to make tor support resolve requests. Need to write
  281. a script and an interface (including an extension to the socks
  282. protocol) so we can ask it to do resolve requests. Need to patch
  283. tsocks to intercept gethostbyname, else we'll continue leaking it.
  284. Improve path selection algorithms based on routing-zones paper. Be sure
  285. to start and end circuits in different ASs. Ideally, consider AS of
  286. source and destination -- maybe even enter and exit via nearby AS.
  287. Intermediate model, with some delays and mixing.
  288. Add defensive dropping regime?
  289. Make it more correct:
  290. Handle half-open connections: right now we don't support all TCP
  291. streams, at least according to the protocol. But we handle all that
  292. we've seen in the wild.
  293. Support IPv6.
  294. Efficiency/speed/robustness:
  295. Congestion control. Is our current design sufficient once we have heavy
  296. use? Need to measure and tweak, or maybe overhaul.
  297. Allow small cells and large cells on the same network?
  298. Cell buffering and resending. This will allow us to handle broken
  299. circuits as long as the endpoints don't break, plus will allow
  300. connection (tls session key) rotation.
  301. Implement Morphmix, so we can compare its behavior, complexity, etc.
  302. Use cpuworker for more heavy lifting.
  303. - Signing (and verifying) hidserv descriptors
  304. - Signing (and verifying) intro/rend requests
  305. - Signing (and verifying) router descriptors
  306. - Signing (and verifying) directories
  307. - Doing TLS handshake (this is very hard to separate out, though)
  308. Buffer size pool: allocate a maximum size for all buffers, not
  309. a maximum size for each buffer. So we don't have to give up as
  310. quickly (and kill the thickpipe!) when there's congestion.
  311. Exit node caching: tie into squid or other caching web proxy.
  312. Other transport. HTTP, udp, rdp, airhook, etc. May have to do our own
  313. link crypto, unless we can bully openssl into it.
  314. P2P Tor:
  315. Do all the scalability stuff above, first.
  316. Incentives to relay. Not so hard.
  317. Incentives to allow exit. Possibly quite hard.
  318. Sybil defenses without having a human bottleneck.
  319. How to gather random sample of nodes.
  320. How to handle nodelist recommendations.
  321. Consider incremental switches: a p2p tor with only 50 users has
  322. different anonymity properties than one with 10k users, and should
  323. be treated differently.