TODO 13 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269
  1. mutiny suggests: if none of the ports is defined maybe it shouldn't start.
  2. aaron got a crash in tor_timegm in tzset on os x, with -l warn but not with -l debug.
  3. Oct 25 04:29:17.017 [warn] directory_initiate_command(): No running dirservers known. This is really bad.
  4. rename ACI to CircID
  5. rotate tls-level connections -- make new ones, expire old ones.
  6. dirserver shouldn't put you in running-routers list if you haven't
  7. uploading a descriptor recently
  8. look at having smallcells and largecells
  9. separate trying to rebuild a circuit because you have none from trying to rebuild a
  10. circuit because the current one is stale
  11. <nickm> If I compromise a node, and streamIDs are sequential, I learn
  12. how many streams have been open and closed on this circuit at this point.
  13. > hm. you learn this for circuits too, do you not?
  14. <nickm> True. But how-many-circuits-from-A-to-B only leaks how long
  15. the connection from A to B has been alive and how much use it's seen.
  16. > ok. needs more investigation.
  17. Legend:
  18. SPEC!! - Not specified
  19. SPEC - Spec not finalized
  20. NICK - nick claims
  21. ARMA - arma claims
  22. - Not done
  23. * Top priority
  24. . Partially done
  25. o Done
  26. D Deferred
  27. X Abandoned
  28. Short-term:
  29. . integrate rep_ok functions, see what breaks
  30. - update tor faq
  31. . obey SocksBindAddress, ORBindAddress
  32. - warn if we're running as root
  33. - make connection_flush_buf() more obviously obsolete
  34. . let hup reread the config file, eg so we can get new exit
  35. policies without restarting
  36. - use times(2) rather than gettimeofday to measure how long it
  37. takes to process a cell
  38. . Exit policies
  39. o Spec how to write the exit policies
  40. - Path selection algorithms
  41. - Let user request certain nodes
  42. - And disallow certain nodes
  43. D Choose path by jurisdiction, etc?
  44. - Make relay end cells have failure status and payload attached
  45. - Streams that fail due to exit policy must reextend to new node
  46. - Add extend_wait state to edge connections, thumb through them
  47. when the AP get an extended cell.
  48. - let non-approved routers handshake.
  49. - just list approved routers in directory.
  50. . migrate to using nickname rather than addr:port for routers
  51. o decide_aci_type
  52. - generate onion skins
  53. - circuit_send_next_onion_skin
  54. - circuit_extend
  55. - onion_generate_cpath
  56. - get_unique_aci_by_addr_port
  57. - circ->n_addr and circ->n_port
  58. - circuit_enumerate_by_naddr_nport
  59. - cpath layers
  60. - connection_or_connect
  61. - connection_exact_get_by_addr_port
  62. - connection_twin_get_by_addr_port
  63. - router_get_by_addr_port
  64. - connection_or_init_conn_from_router
  65. - tag_pack, tag_unpack, connection_cpu_process_inbuf
  66. - directory_initiate_command
  67. . Move from onions to ephemeral DH
  68. o incremental path building
  69. o transition circuit-level sendmes to hop-level sendmes
  70. o implement truncate, truncated
  71. o move from 192byte DH to 128byte DH, so it isn't so damn slow
  72. - exiting from not-last hop
  73. - OP logic to decide to extend/truncate a path
  74. - make sure exiting from the not-last hop works
  75. - logic to find last *open* hop, not last hop, in cpath
  76. - choose exit nodes by exit policies
  77. - Remember address and port when resolving.
  78. - Extend by nickname/hostname/something, not by IP.
  79. On-going
  80. . Better comments for functions!
  81. . Go through log messages, reduce confusing error messages.
  82. . make the logs include more info (fd, etc)
  83. . Unit tests
  84. Mid-term:
  85. . Redo scheduler
  86. o fix SSL_read bug for buffered records
  87. - make round-robining more fair
  88. - What happens when a circuit's length is 1? What breaks?
  89. . streams / circuits
  90. o Implement streams
  91. o Rotate circuits after N minutes?
  92. X Circuits should expire when circuit->expire triggers
  93. NICK . Handle half-open connections
  94. o openssh is an application that uses half-open connections
  95. o Figure out what causes connections to close, standardize
  96. when we mark a connection vs when we tear it down
  97. o Look at what ssl does to keep from mutating data streams
  98. ARMA - Reduce streamid footprint from 7 bytes to 2 bytes
  99. - Check for collisions in streamid (now possible with
  100. just 2 bytes), and back up & replace with padding if so
  101. - Use the 3 saved bytes to put pseudorandomness in each relay cell
  102. - Use the 4 reserved bytes in each cell header to keep 1/5
  103. of a sha1 of the relay payload (move into stream header)
  104. - Move length into the stream header too
  105. - Spec the stream_id stuff. Clarify that nobody on the backward
  106. stream should look at stream_id.
  107. . Put CPU workers in separate processes
  108. o Handle multiple cpu workers (one for each cpu, plus one)
  109. o Queue for pending tasks if all workers full
  110. o Support the 'process this onion' task
  111. D Merge dnsworkers and cpuworkers to some extent
  112. - Handle cpuworkers dying
  113. . Scrubbing proxies
  114. - Find an smtp proxy?
  115. - Check the old smtp proxy code
  116. o Find an ftp proxy? wget --passive
  117. D Wait until there are packet redirectors for Linux
  118. . Get socks4a support into Mozilla
  119. . Develop rendezvous points
  120. SPEC!! - Handle socks commands other than connect, eg, bind?
  121. o Design
  122. - Spec
  123. - Implement
  124. - Tests
  125. o Testing harness/infrastructure
  126. D System tests (how?)
  127. - Performance tests, so we know when we've improved
  128. . webload infrastructure (Bruce)
  129. . httperf infrastructure (easy to set up)
  130. . oprofile (installed in RH >8.0)
  131. NICK . Daemonize and package
  132. o Teach it to fork and background
  133. - Red Hat spec file
  134. - Debian spec file equivalent
  135. . Portability
  136. . Which .h files are we actually using?
  137. . Port to:
  138. o Linux
  139. o BSD
  140. . Solaris
  141. o Cygwin
  142. . Win32
  143. o OS X
  144. - deal with pollhup / reached_eof on all platforms
  145. o openssl randomness
  146. o inet_ntoa
  147. . stdint.h
  148. - Make a script to set up a local network on your machine
  149. - More flexibility in node addressing
  150. D Support IPv6 rather than just 4
  151. - Handle multihomed servers (config variable to set IP)
  152. In the distant future:
  153. D Load balancing between router twins
  154. D Keep track of load over links/nodes, to
  155. know who's hosed
  156. SPEC!! D Non-clique topologies
  157. D Implement our own memory management, at least for common structs
  158. (Not ever necessary?)
  159. D Advanced directory servers
  160. D Automated reputation management
  161. SPEC!! D Figure out how to do threshold directory servers
  162. D jurisdiction info in dirserver entries? other info?
  163. Older (done) todo stuff:
  164. o Get tor to act like a socks server
  165. o socks4, socks4a
  166. o socks5
  167. o routers have identity key, link key, onion key.
  168. o link key certs are
  169. D signed by identity key
  170. D not in descriptor
  171. o not in config
  172. D not on disk
  173. o identity and onion keys are in descriptor (and disk)
  174. o upon boot, if it doesn't find identity key, generate it and write it.
  175. o also write a file with the identity key fingerprint in it
  176. o router generates descriptor: flesh out router_get_my_descriptor()
  177. o Routers sign descriptors with identity key
  178. o routers put version number in descriptor
  179. o routers should maybe have `uname -a` in descriptor?
  180. o Give nicknames to routers
  181. o in config
  182. o in descriptors
  183. o router posts descriptor
  184. o when it boots
  185. D when it changes
  186. o change tls stuff so certs don't get written to disk, or read from disk
  187. o make directory.c 'thread'safe
  188. o dirserver parses descriptor
  189. o dirserver checks signature
  190. D client checks signature?
  191. o dirserver writes directory to file
  192. o reads that file upon boot
  193. o directory includes all routers, up and down
  194. o add "up" line to directory, listing nicknames
  195. o instruments ORs to report stats
  196. o average cell fullness
  197. o average bandwidth used
  198. o configure log files. separate log file, separate severities.
  199. o what assumptions break if we fclose(0) when we daemonize?
  200. o make buffer struct elements opaque outside buffers.c
  201. o add log convention to the HACKING file
  202. o make 'make install' do the right thing
  203. o change binary name to tor
  204. o change config files so you look at commandline, else look in
  205. /etc/torrc. no cascading.
  206. o have an absolute datadir with fixed names for files, and fixed-name
  207. keydir under that with fixed names
  208. o Move (most of) the router/directory code out of main.c
  209. o Simple directory servers
  210. o Include key in source; sign directories
  211. o Signed directory backend
  212. o Document
  213. o Integrate
  214. o Add versions to code
  215. o Have directories list recommended-versions
  216. o Include line in directories
  217. o Check for presence of line.
  218. o Quit if running the wrong version
  219. o Command-line option to override quit
  220. o Add more information to directory server entries
  221. o Exit policies
  222. o Clearer bandwidth management
  223. o Do we want to remove bandwidth from OR handshakes?
  224. o What about OP handshakes?
  225. X Move away from openssl
  226. o Abstract out crypto calls
  227. X Look at nss, others? Just include code?
  228. o Use a stronger cipher
  229. o aes now, by including the code ourselves
  230. X On the fly compression of each stream
  231. o Clean up the event loop (optimize and sanitize)
  232. o Remove that awful concept of 'roles'
  233. o Terminology
  234. o Circuits, topics, cells stay named that
  235. o 'Connection' gets divided, or renamed, or something?
  236. o DNS farm
  237. o Distribute queries onto the farm, get answers
  238. o Preemptively grow a new worker before he's needed
  239. o Prune workers when too many are idle
  240. o DNS cache
  241. o Clear DNS cache over time
  242. D Honor DNS TTL info (how??)
  243. o Have strategy when all workers are busy
  244. o Keep track of which connections are in dns_wait
  245. o Need to cache positives/negatives on the tor side
  246. o Keep track of which queries have been asked
  247. o Better error handling when
  248. o An address doesn't resolve
  249. o We have max workers running
  250. o Consider taking the master out of the loop?
  251. X Implement reply onions
  252. o Total rate limiting
  253. o Look at OR handshake in more detail
  254. o Spec it
  255. o Merge OR and OP handshakes
  256. o rearrange connection_or so it doesn't suck so much to read
  257. D Periodic link key rotation. Spec?
  258. o wrap malloc with something that explodes when it fails
  259. o Clean up the number of places that get to look at prkey