TODO 15 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320
  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. - on solaris, need to build with
  14. LDFLAGS="-lsocket -lnsl" ./configure --with-ssl-dir=/usr/local/ssl
  15. - should maybe make clients exit(1) when bad things happen?
  16. e.g. clock skew.
  17. o client-side dns cache doesn't appear to be getting populated
  18. by 'connected' cells. In fact, the 'connected' cells don't even
  19. include the IP.
  20. o When it can't resolve any dirservers, it is useless from then on.
  21. We should make it reload the RouterFile if it has no dirservers.
  22. o Sometimes it picks a middleman node as the exit for a circuit.
  23. o if you specify a non-dirserver as exitnode or entrynode, when it
  24. makes the first few circuits it hasn't yet fetched the directory,
  25. so it warns that it doesn't know the node.
  26. o make 'make test' exit(1) if a test fails.
  27. . fix buffer unit test so it passes
  28. Short-term:
  29. o put IP into the descriptor, so clients don't need to resolve things
  30. o when you hup, rewrite the router.desc file (and maybe others)
  31. - consider handling broken socks4 implementations
  32. - improve how it behaves when i remove a line from the approved-routers files
  33. - Make tls connections tls_close intentionally
  34. - Add '[...truncated]' or similar to truncated log entries (like the directory
  35. in connection_dir_process_inbuf()).
  36. - Make logs handle it better when writing to them fails.
  37. - leave server descriptor out of directory if it's too old
  38. o Rename ACI to circID
  39. . integrate rep_ok functions, see what breaks
  40. - update tor faq
  41. o obey SocksBindAddress, ORBindAddress
  42. o warn if we're running as root
  43. o make connection_flush_buf() more obviously obsolete
  44. o let hup reread the config file, eg so we can get new exit
  45. policies without restarting
  46. o Put recommended_versions in a config entry
  47. X use times(2) rather than gettimeofday to measure how long it
  48. takes to process a cell
  49. o Separate trying to rebuild a circuit because you have none from trying
  50. to rebuild a circuit because the current one is stale
  51. X Continue reading from socks port even while waiting for connect.
  52. o Exit policies
  53. o Spec how to write the exit policies
  54. o Path selection algorithms
  55. o Choose path more incrementally
  56. o Let user request first/last node
  57. o And disallow certain nodes
  58. D Choose path by jurisdiction, etc?
  59. o Make relay end cells have failure status and payload attached
  60. X let non-approved routers handshake.
  61. X Dirserver shouldn't put you in running-routers list if you haven't
  62. uploaded a descriptor recently
  63. X migrate to using nickname rather than addr:port for routers
  64. - migrate to using IPv6 sizes everywhere
  65. o Move from onions to ephemeral DH
  66. o incremental path building
  67. o transition circuit-level sendmes to hop-level sendmes
  68. o implement truncate, truncated
  69. o move from 192byte DH to 128byte DH, so it isn't so damn slow
  70. X exiting from not-last hop
  71. X OP logic to decide to extend/truncate a path
  72. X make sure exiting from the not-last hop works
  73. X logic to find last *open* hop, not last hop, in cpath
  74. o Remember address and port when beginning.
  75. - Extend by nickname/hostname/something, not by IP.
  76. - Need a relay teardown cell, separate from one-way ends.
  77. X remove per-connection rate limiting
  78. - Make it harder to circumvent bandwidth caps: look at number of bytes
  79. sent across sockets, not number sent inside TLS stream.
  80. On-going
  81. . Better comments for functions!
  82. . Go through log messages, reduce confusing error messages.
  83. . make the logs include more info (fd, etc)
  84. . Unit tests
  85. . Update the spec so it matches the code
  86. Mid-term:
  87. - Rotate tls-level connections -- make new ones, expire old ones.
  88. So we get actual key rotation, not just symmetric key rotation
  89. o Are there anonymity issues with sequential streamIDs? Sequential
  90. circIDs? Eg an attacker can learn how many there have been.
  91. The fix is to initialize them randomly rather than at 1.
  92. - Look at having smallcells and largecells
  93. . Redo scheduler
  94. o fix SSL_read bug for buffered records
  95. - make round-robining more fair
  96. - What happens when a circuit's length is 1? What breaks?
  97. . streams / circuits
  98. o Implement streams
  99. o Rotate circuits after N minutes?
  100. X Circuits should expire when circuit->expire triggers
  101. NICK . Handle half-open connections
  102. o openssh is an application that uses half-open connections
  103. o Figure out what causes connections to close, standardize
  104. when we mark a connection vs when we tear it down
  105. o Look at what ssl does to keep from mutating data streams
  106. o Put CPU workers in separate processes
  107. o Handle multiple cpu workers (one for each cpu, plus one)
  108. o Queue for pending tasks if all workers full
  109. o Support the 'process this onion' task
  110. D Merge dnsworkers and cpuworkers to some extent
  111. o Handle cpuworkers dying
  112. . Scrubbing proxies
  113. - Find an smtp proxy?
  114. - Check the old smtp proxy code
  115. o Find an ftp proxy? wget --passive
  116. D Wait until there are packet redirectors for Linux
  117. . Get socks4a support into Mozilla
  118. . Develop rendezvous points
  119. X Handle socks commands other than connect, eg, bind?
  120. o Design
  121. - Spec
  122. - Implement
  123. . Tests
  124. o Testing harness/infrastructure
  125. D System tests (how?)
  126. - Performance tests, so we know when we've improved
  127. . webload infrastructure (Bruce)
  128. . httperf infrastructure (easy to set up)
  129. . oprofile (installed in RH >8.0)
  130. NICK . Daemonize and package
  131. o Teach it to fork and background
  132. . Red Hat spec file
  133. o Debian spec file equivalent
  134. . Portability
  135. . Which .h files are we actually using?
  136. . Port to:
  137. o Linux
  138. o BSD
  139. . Solaris
  140. o Cygwin
  141. . Win32
  142. o OS X
  143. - deal with pollhup / reached_eof on all platforms
  144. o openssl randomness
  145. o inet_ntoa
  146. o stdint.h
  147. - Make a script to set up a local network on your machine
  148. o More flexibility in node addressing
  149. D Support IPv6 rather than just 4
  150. o Handle multihomed servers (config variable to set IP)
  151. In the distant future:
  152. D Load balancing between router twins
  153. D Keep track of load over links/nodes, to
  154. know who's hosed
  155. SPEC!! D Non-clique topologies
  156. D Implement our own memory management, at least for common structs
  157. (Not ever necessary?)
  158. D Advanced directory servers
  159. D Automated reputation management
  160. SPEC!! D Figure out how to do threshold directory servers
  161. D jurisdiction info in dirserver entries? other info?
  162. Older (done) todo stuff:
  163. For 0.0.2pre17:
  164. o Put a H(K | handshake) into the onionskin response
  165. o Make cells 512 bytes
  166. o Reduce streamid footprint from 7 bytes to 2 bytes
  167. X Check for collisions in streamid (now possible with
  168. just 2 bytes), and back up & replace with padding if so
  169. o Use the 4 reserved bytes in each cell header to keep 1/5
  170. of a sha1 of the ongoing relay payload (move into stream header)
  171. o Move length into the stream header too
  172. o Make length 2 bytes
  173. D increase DH key length
  174. D increase RSA key length
  175. D Spec the stream_id stuff. Clarify that nobody on the backward
  176. stream should look at stream_id.
  177. Cell:
  178. ACI (anonymous circuit identifier) [2 bytes]
  179. Command [1 byte]
  180. Payload (padded with 0 bytes) [509 bytes]
  181. Relay payload:
  182. Relay command [1 byte]
  183. Stream ID [7 bytes]
  184. Partial SHA-1 [4 bytes]
  185. Length [2 bytes]
  186. Relay payload [495 bytes]
  187. For 0.0.2pre15:
  188. o don't pick exit nodes which will certainly reject all things.
  189. o don't pick nodes that the directory says are down
  190. o choose randomly from running dirservers, not just first one
  191. o install the man page
  192. o warn when client-side tries an address/port which no router in the dir accepts.
  193. For 0.0.2pre14:
  194. o More flexible exit policies (18.*, 18.0.0.0/8)
  195. o Work to succeed in the precense of exit policy violation
  196. o Replace desired_path_len with opaque path-selection specifier
  197. o Client-side DNS caching
  198. o Add entries to client DNS cache based on END cells
  199. o Remove port from END_REASON_EXITPOLICY cells
  200. o Start building new circuits when we get an exit-policy
  201. failure. (Defer exiting from the middle of existing
  202. circuits or extending existing circuits for later.)
  203. o Implement function to check whether a routerinfo_t
  204. supports a given exit addr.
  205. o Choose the exit node of an in-progress circuit based on
  206. pending AP connections.
  207. o Choose the exit node _first_, then beginning, then
  208. middle nodes.
  209. Previous:
  210. o Get tor to act like a socks server
  211. o socks4, socks4a
  212. o socks5
  213. o routers have identity key, link key, onion key.
  214. o link key certs are
  215. D signed by identity key
  216. D not in descriptor
  217. o not in config
  218. D not on disk
  219. o identity and onion keys are in descriptor (and disk)
  220. o upon boot, if it doesn't find identity key, generate it and write it.
  221. o also write a file with the identity key fingerprint in it
  222. o router generates descriptor: flesh out router_get_my_descriptor()
  223. o Routers sign descriptors with identity key
  224. o routers put version number in descriptor
  225. o routers should maybe have `uname -a` in descriptor?
  226. o Give nicknames to routers
  227. o in config
  228. o in descriptors
  229. o router posts descriptor
  230. o when it boots
  231. o every DirFetchPostPeriod seconds
  232. D when it changes
  233. o change tls stuff so certs don't get written to disk, or read from disk
  234. o make directory.c 'thread'safe
  235. o dirserver parses descriptor
  236. o dirserver checks signature
  237. D client checks signature?
  238. o dirserver writes directory to file
  239. o reads that file upon boot
  240. o directory includes all routers, up and down
  241. o add "up" line to directory, listing nicknames
  242. o instruments ORs to report stats
  243. o average cell fullness
  244. o average bandwidth used
  245. o configure log files. separate log file, separate severities.
  246. o what assumptions break if we fclose(0) when we daemonize?
  247. o make buffer struct elements opaque outside buffers.c
  248. o add log convention to the HACKING file
  249. o make 'make install' do the right thing
  250. o change binary name to tor
  251. o change config files so you look at commandline, else look in
  252. /etc/torrc. no cascading.
  253. o have an absolute datadir with fixed names for files, and fixed-name
  254. keydir under that with fixed names
  255. o Move (most of) the router/directory code out of main.c
  256. o Simple directory servers
  257. o Include key in source; sign directories
  258. o Signed directory backend
  259. o Document
  260. o Integrate
  261. o Add versions to code
  262. o Have directories list recommended-versions
  263. o Include line in directories
  264. o Check for presence of line.
  265. o Quit if running the wrong version
  266. o Command-line option to override quit
  267. o Add more information to directory server entries
  268. o Exit policies
  269. o Clearer bandwidth management
  270. o Do we want to remove bandwidth from OR handshakes?
  271. o What about OP handshakes?
  272. X Move away from openssl
  273. o Abstract out crypto calls
  274. X Look at nss, others? Just include code?
  275. o Use a stronger cipher
  276. o aes now, by including the code ourselves
  277. X On the fly compression of each stream
  278. o Clean up the event loop (optimize and sanitize)
  279. o Remove that awful concept of 'roles'
  280. o Terminology
  281. o Circuits, topics, cells stay named that
  282. o 'Connection' gets divided, or renamed, or something?
  283. o DNS farm
  284. o Distribute queries onto the farm, get answers
  285. o Preemptively grow a new worker before he's needed
  286. o Prune workers when too many are idle
  287. o DNS cache
  288. o Clear DNS cache over time
  289. D Honor DNS TTL info (how??)
  290. o Have strategy when all workers are busy
  291. o Keep track of which connections are in dns_wait
  292. o Need to cache positives/negatives on the tor side
  293. o Keep track of which queries have been asked
  294. o Better error handling when
  295. o An address doesn't resolve
  296. o We have max workers running
  297. o Consider taking the master out of the loop?
  298. X Implement reply onions
  299. o Total rate limiting
  300. o Look at OR handshake in more detail
  301. o Spec it
  302. o Merge OR and OP handshakes
  303. o rearrange connection_or so it doesn't suck so much to read
  304. D Periodic link key rotation. Spec?
  305. o wrap malloc with something that explodes when it fails
  306. o Clean up the number of places that get to look at prkey