TODO 15 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318
  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. Bugs:
  13. - should maybe make clients exit(1) when bad things happen?
  14. e.g. clock skew.
  15. o client-side dns cache doesn't appear to be getting populated
  16. by 'connected' cells. In fact, the 'connected' cells don't even
  17. include the IP.
  18. o When it can't resolve any dirservers, it is useless from then on.
  19. We should make it reload the RouterFile if it has no dirservers.
  20. o Sometimes it picks a middleman node as the exit for a circuit.
  21. o if you specify a non-dirserver as exitnode or entrynode, when it
  22. makes the first few circuits it hasn't yet fetched the directory,
  23. so it warns that it doesn't know the node.
  24. o make 'make test' exit(1) if a test fails.
  25. . fix buffer unit test so it passes
  26. Short-term:
  27. o put IP into the descriptor, so clients don't need to resolve things
  28. o when you hup, rewrite the router.desc file (and maybe others)
  29. - consider handling broken socks4 implementations
  30. - improve how it behaves when i remove a line from the approved-routers files
  31. - Make tls connections tls_close intentionally
  32. - Add '[...truncated]' or similar to truncated log entries (like the directory
  33. in connection_dir_process_inbuf()).
  34. - Make logs handle it better when writing to them fails.
  35. - leave server descriptor out of directory if it's too old
  36. o Rename ACI to circID
  37. . integrate rep_ok functions, see what breaks
  38. - update tor faq
  39. o obey SocksBindAddress, ORBindAddress
  40. o warn if we're running as root
  41. o make connection_flush_buf() more obviously obsolete
  42. o let hup reread the config file, eg so we can get new exit
  43. policies without restarting
  44. o Put recommended_versions in a config entry
  45. X use times(2) rather than gettimeofday to measure how long it
  46. takes to process a cell
  47. o Separate trying to rebuild a circuit because you have none from trying
  48. to rebuild a circuit because the current one is stale
  49. X Continue reading from socks port even while waiting for connect.
  50. o Exit policies
  51. o Spec how to write the exit policies
  52. o Path selection algorithms
  53. o Choose path more incrementally
  54. o Let user request first/last node
  55. o And disallow certain nodes
  56. D Choose path by jurisdiction, etc?
  57. o Make relay end cells have failure status and payload attached
  58. X let non-approved routers handshake.
  59. X Dirserver shouldn't put you in running-routers list if you haven't
  60. uploaded a descriptor recently
  61. X migrate to using nickname rather than addr:port for routers
  62. - migrate to using IPv6 sizes everywhere
  63. o Move from onions to ephemeral DH
  64. o incremental path building
  65. o transition circuit-level sendmes to hop-level sendmes
  66. o implement truncate, truncated
  67. o move from 192byte DH to 128byte DH, so it isn't so damn slow
  68. X exiting from not-last hop
  69. X OP logic to decide to extend/truncate a path
  70. X make sure exiting from the not-last hop works
  71. X logic to find last *open* hop, not last hop, in cpath
  72. o Remember address and port when beginning.
  73. - Extend by nickname/hostname/something, not by IP.
  74. - Need a relay teardown cell, separate from one-way ends.
  75. X remove per-connection rate limiting
  76. - Make it harder to circumvent bandwidth caps: look at number of bytes
  77. sent across sockets, not number sent inside TLS stream.
  78. On-going
  79. . Better comments for functions!
  80. . Go through log messages, reduce confusing error messages.
  81. . make the logs include more info (fd, etc)
  82. . Unit tests
  83. . Update the spec so it matches the code
  84. Mid-term:
  85. - Rotate tls-level connections -- make new ones, expire old ones.
  86. So we get actual key rotation, not just symmetric key rotation
  87. o Are there anonymity issues with sequential streamIDs? Sequential
  88. circIDs? Eg an attacker can learn how many there have been.
  89. The fix is to initialize them randomly rather than at 1.
  90. - Look at having smallcells and largecells
  91. . Redo scheduler
  92. o fix SSL_read bug for buffered records
  93. - make round-robining more fair
  94. - What happens when a circuit's length is 1? What breaks?
  95. . streams / circuits
  96. o Implement streams
  97. o Rotate circuits after N minutes?
  98. X Circuits should expire when circuit->expire triggers
  99. NICK . Handle half-open connections
  100. o openssh is an application that uses half-open connections
  101. o Figure out what causes connections to close, standardize
  102. when we mark a connection vs when we tear it down
  103. o Look at what ssl does to keep from mutating data streams
  104. o Put CPU workers in separate processes
  105. o Handle multiple cpu workers (one for each cpu, plus one)
  106. o Queue for pending tasks if all workers full
  107. o Support the 'process this onion' task
  108. D Merge dnsworkers and cpuworkers to some extent
  109. o Handle cpuworkers dying
  110. . Scrubbing proxies
  111. - Find an smtp proxy?
  112. - Check the old smtp proxy code
  113. o Find an ftp proxy? wget --passive
  114. D Wait until there are packet redirectors for Linux
  115. . Get socks4a support into Mozilla
  116. . Develop rendezvous points
  117. X Handle socks commands other than connect, eg, bind?
  118. o Design
  119. - Spec
  120. - Implement
  121. . Tests
  122. o Testing harness/infrastructure
  123. D System tests (how?)
  124. - Performance tests, so we know when we've improved
  125. . webload infrastructure (Bruce)
  126. . httperf infrastructure (easy to set up)
  127. . oprofile (installed in RH >8.0)
  128. NICK . Daemonize and package
  129. o Teach it to fork and background
  130. . Red Hat spec file
  131. o Debian spec file equivalent
  132. . Portability
  133. . Which .h files are we actually using?
  134. . Port to:
  135. o Linux
  136. o BSD
  137. . Solaris
  138. o Cygwin
  139. . Win32
  140. o OS X
  141. - deal with pollhup / reached_eof on all platforms
  142. o openssl randomness
  143. o inet_ntoa
  144. o stdint.h
  145. - Make a script to set up a local network on your machine
  146. o More flexibility in node addressing
  147. D Support IPv6 rather than just 4
  148. o Handle multihomed servers (config variable to set IP)
  149. In the distant future:
  150. D Load balancing between router twins
  151. D Keep track of load over links/nodes, to
  152. know who's hosed
  153. SPEC!! D Non-clique topologies
  154. D Implement our own memory management, at least for common structs
  155. (Not ever necessary?)
  156. D Advanced directory servers
  157. D Automated reputation management
  158. SPEC!! D Figure out how to do threshold directory servers
  159. D jurisdiction info in dirserver entries? other info?
  160. Older (done) todo stuff:
  161. For 0.0.2pre17:
  162. o Put a H(K | handshake) into the onionskin response
  163. o Make cells 512 bytes
  164. o Reduce streamid footprint from 7 bytes to 2 bytes
  165. X Check for collisions in streamid (now possible with
  166. just 2 bytes), and back up & replace with padding if so
  167. o Use the 4 reserved bytes in each cell header to keep 1/5
  168. of a sha1 of the ongoing relay payload (move into stream header)
  169. o Move length into the stream header too
  170. o Make length 2 bytes
  171. D increase DH key length
  172. D increase RSA key length
  173. D Spec the stream_id stuff. Clarify that nobody on the backward
  174. stream should look at stream_id.
  175. Cell:
  176. ACI (anonymous circuit identifier) [2 bytes]
  177. Command [1 byte]
  178. Payload (padded with 0 bytes) [509 bytes]
  179. Relay payload:
  180. Relay command [1 byte]
  181. Stream ID [7 bytes]
  182. Partial SHA-1 [4 bytes]
  183. Length [2 bytes]
  184. Relay payload [495 bytes]
  185. For 0.0.2pre15:
  186. o don't pick exit nodes which will certainly reject all things.
  187. o don't pick nodes that the directory says are down
  188. o choose randomly from running dirservers, not just first one
  189. o install the man page
  190. o warn when client-side tries an address/port which no router in the dir accepts.
  191. For 0.0.2pre14:
  192. o More flexible exit policies (18.*, 18.0.0.0/8)
  193. o Work to succeed in the precense of exit policy violation
  194. o Replace desired_path_len with opaque path-selection specifier
  195. o Client-side DNS caching
  196. o Add entries to client DNS cache based on END cells
  197. o Remove port from END_REASON_EXITPOLICY cells
  198. o Start building new circuits when we get an exit-policy
  199. failure. (Defer exiting from the middle of existing
  200. circuits or extending existing circuits for later.)
  201. o Implement function to check whether a routerinfo_t
  202. supports a given exit addr.
  203. o Choose the exit node of an in-progress circuit based on
  204. pending AP connections.
  205. o Choose the exit node _first_, then beginning, then
  206. middle nodes.
  207. Previous:
  208. o Get tor to act like a socks server
  209. o socks4, socks4a
  210. o socks5
  211. o routers have identity key, link key, onion key.
  212. o link key certs are
  213. D signed by identity key
  214. D not in descriptor
  215. o not in config
  216. D not on disk
  217. o identity and onion keys are in descriptor (and disk)
  218. o upon boot, if it doesn't find identity key, generate it and write it.
  219. o also write a file with the identity key fingerprint in it
  220. o router generates descriptor: flesh out router_get_my_descriptor()
  221. o Routers sign descriptors with identity key
  222. o routers put version number in descriptor
  223. o routers should maybe have `uname -a` in descriptor?
  224. o Give nicknames to routers
  225. o in config
  226. o in descriptors
  227. o router posts descriptor
  228. o when it boots
  229. o every DirFetchPostPeriod seconds
  230. D when it changes
  231. o change tls stuff so certs don't get written to disk, or read from disk
  232. o make directory.c 'thread'safe
  233. o dirserver parses descriptor
  234. o dirserver checks signature
  235. D client checks signature?
  236. o dirserver writes directory to file
  237. o reads that file upon boot
  238. o directory includes all routers, up and down
  239. o add "up" line to directory, listing nicknames
  240. o instruments ORs to report stats
  241. o average cell fullness
  242. o average bandwidth used
  243. o configure log files. separate log file, separate severities.
  244. o what assumptions break if we fclose(0) when we daemonize?
  245. o make buffer struct elements opaque outside buffers.c
  246. o add log convention to the HACKING file
  247. o make 'make install' do the right thing
  248. o change binary name to tor
  249. o change config files so you look at commandline, else look in
  250. /etc/torrc. no cascading.
  251. o have an absolute datadir with fixed names for files, and fixed-name
  252. keydir under that with fixed names
  253. o Move (most of) the router/directory code out of main.c
  254. o Simple directory servers
  255. o Include key in source; sign directories
  256. o Signed directory backend
  257. o Document
  258. o Integrate
  259. o Add versions to code
  260. o Have directories list recommended-versions
  261. o Include line in directories
  262. o Check for presence of line.
  263. o Quit if running the wrong version
  264. o Command-line option to override quit
  265. o Add more information to directory server entries
  266. o Exit policies
  267. o Clearer bandwidth management
  268. o Do we want to remove bandwidth from OR handshakes?
  269. o What about OP handshakes?
  270. X Move away from openssl
  271. o Abstract out crypto calls
  272. X Look at nss, others? Just include code?
  273. o Use a stronger cipher
  274. o aes now, by including the code ourselves
  275. X On the fly compression of each stream
  276. o Clean up the event loop (optimize and sanitize)
  277. o Remove that awful concept of 'roles'
  278. o Terminology
  279. o Circuits, topics, cells stay named that
  280. o 'Connection' gets divided, or renamed, or something?
  281. o DNS farm
  282. o Distribute queries onto the farm, get answers
  283. o Preemptively grow a new worker before he's needed
  284. o Prune workers when too many are idle
  285. o DNS cache
  286. o Clear DNS cache over time
  287. D Honor DNS TTL info (how??)
  288. o Have strategy when all workers are busy
  289. o Keep track of which connections are in dns_wait
  290. o Need to cache positives/negatives on the tor side
  291. o Keep track of which queries have been asked
  292. o Better error handling when
  293. o An address doesn't resolve
  294. o We have max workers running
  295. o Consider taking the master out of the loop?
  296. X Implement reply onions
  297. o Total rate limiting
  298. o Look at OR handshake in more detail
  299. o Spec it
  300. o Merge OR and OP handshakes
  301. o rearrange connection_or so it doesn't suck so much to read
  302. D Periodic link key rotation. Spec?
  303. o wrap malloc with something that explodes when it fails
  304. o Clean up the number of places that get to look at prkey