TODO 15 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327
  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. o we call signal(), but we should be calling sigaction()
  14. - send socks rejects when things go bad ?
  15. o on solaris, need to build with
  16. LDFLAGS="-lsocket -lnsl" ./configure
  17. o on solaris, we HAVE_UNAME but the uname() call fails?
  18. - should maybe make clients exit(1) when bad things happen?
  19. e.g. clock skew.
  20. o client-side dns cache doesn't appear to be getting populated
  21. by 'connected' cells. In fact, the 'connected' cells don't even
  22. include the IP.
  23. o When it can't resolve any dirservers, it is useless from then on.
  24. We should make it reload the RouterFile if it has no dirservers.
  25. o Sometimes it picks a middleman node as the exit for a circuit.
  26. o if you specify a non-dirserver as exitnode or entrynode, when it
  27. makes the first few circuits it hasn't yet fetched the directory,
  28. so it warns that it doesn't know the node.
  29. o make 'make test' exit(1) if a test fails.
  30. . fix buffer unit test so it passes
  31. Short-term:
  32. - make recommendedversions different for clients and servers.
  33. e.g. C0.0.3 vs S0.0.3?
  34. o put IP into the descriptor, so clients don't need to resolve things
  35. o when you hup, rewrite the router.desc file (and maybe others)
  36. - consider handling broken socks4 implementations
  37. - improve how it behaves when i remove a line from the approved-routers files
  38. - Make tls connections tls_close intentionally
  39. - Add '[...truncated]' or similar to truncated log entries (like the directory
  40. in connection_dir_process_inbuf()).
  41. - Make logs handle it better when writing to them fails.
  42. - leave server descriptor out of directory if it's too old
  43. o Rename ACI to circID
  44. . integrate rep_ok functions, see what breaks
  45. - update tor faq
  46. o obey SocksBindAddress, ORBindAddress
  47. o warn if we're running as root
  48. o make connection_flush_buf() more obviously obsolete
  49. o let hup reread the config file, eg so we can get new exit
  50. policies without restarting
  51. o Put recommended_versions in a config entry
  52. X use times(2) rather than gettimeofday to measure how long it
  53. takes to process a cell
  54. o Separate trying to rebuild a circuit because you have none from trying
  55. to rebuild a circuit because the current one is stale
  56. X Continue reading from socks port even while waiting for connect.
  57. o Exit policies
  58. o Spec how to write the exit policies
  59. o Path selection algorithms
  60. o Choose path more incrementally
  61. o Let user request first/last node
  62. o And disallow certain nodes
  63. D Choose path by jurisdiction, etc?
  64. o Make relay end cells have failure status and payload attached
  65. X let non-approved routers handshake.
  66. X Dirserver shouldn't put you in running-routers list if you haven't
  67. uploaded a descriptor recently
  68. X migrate to using nickname rather than addr:port for routers
  69. - migrate to using IPv6 sizes everywhere
  70. o Move from onions to ephemeral DH
  71. o incremental path building
  72. o transition circuit-level sendmes to hop-level sendmes
  73. o implement truncate, truncated
  74. o move from 192byte DH to 128byte DH, so it isn't so damn slow
  75. X exiting from not-last hop
  76. X OP logic to decide to extend/truncate a path
  77. X make sure exiting from the not-last hop works
  78. X logic to find last *open* hop, not last hop, in cpath
  79. o Remember address and port when beginning.
  80. - Extend by nickname/hostname/something, not by IP.
  81. - Need a relay teardown cell, separate from one-way ends.
  82. X remove per-connection rate limiting
  83. - Make it harder to circumvent bandwidth caps: look at number of bytes
  84. sent across sockets, not number sent inside TLS stream.
  85. o Audit users of connnection_remove and connection_free outside of
  86. main.c; many should use mark_for_close instead.
  87. On-going
  88. . Better comments for functions!
  89. . Go through log messages, reduce confusing error messages.
  90. . make the logs include more info (fd, etc)
  91. . Unit tests
  92. . Update the spec so it matches the code
  93. Mid-term:
  94. - Rotate tls-level connections -- make new ones, expire old ones.
  95. So we get actual key rotation, not just symmetric key rotation
  96. o Are there anonymity issues with sequential streamIDs? Sequential
  97. circIDs? Eg an attacker can learn how many there have been.
  98. The fix is to initialize them randomly rather than at 1.
  99. - Look at having smallcells and largecells
  100. . Redo scheduler
  101. o fix SSL_read bug for buffered records
  102. - make round-robining more fair
  103. - What happens when a circuit's length is 1? What breaks?
  104. . streams / circuits
  105. o Implement streams
  106. o Rotate circuits after N minutes?
  107. X Circuits should expire when circuit->expire triggers
  108. NICK . Handle half-open connections
  109. o openssh is an application that uses half-open connections
  110. o Figure out what causes connections to close, standardize
  111. when we mark a connection vs when we tear it down
  112. o Look at what ssl does to keep from mutating data streams
  113. o Put CPU workers in separate processes
  114. o Handle multiple cpu workers (one for each cpu, plus one)
  115. o Queue for pending tasks if all workers full
  116. o Support the 'process this onion' task
  117. D Merge dnsworkers and cpuworkers to some extent
  118. o Handle cpuworkers dying
  119. . Scrubbing proxies
  120. - Find an smtp proxy?
  121. - Check the old smtp proxy code
  122. o Find an ftp proxy? wget --passive
  123. D Wait until there are packet redirectors for Linux
  124. . Get socks4a support into Mozilla
  125. . Develop rendezvous points
  126. X Handle socks commands other than connect, eg, bind?
  127. o Design
  128. - Spec
  129. - Implement
  130. . Tests
  131. o Testing harness/infrastructure
  132. D System tests (how?)
  133. - Performance tests, so we know when we've improved
  134. . webload infrastructure (Bruce)
  135. . httperf infrastructure (easy to set up)
  136. . oprofile (installed in RH >8.0)
  137. NICK . Daemonize and package
  138. o Teach it to fork and background
  139. . Red Hat spec file
  140. o Debian spec file equivalent
  141. . Portability
  142. . Which .h files are we actually using?
  143. . Port to:
  144. o Linux
  145. o BSD
  146. . Solaris
  147. o Cygwin
  148. . Win32
  149. o OS X
  150. - deal with pollhup / reached_eof on all platforms
  151. o openssl randomness
  152. o inet_ntoa
  153. o stdint.h
  154. - Make a script to set up a local network on your machine
  155. o More flexibility in node addressing
  156. D Support IPv6 rather than just 4
  157. o Handle multihomed servers (config variable to set IP)
  158. In the distant future:
  159. D Load balancing between router twins
  160. D Keep track of load over links/nodes, to
  161. know who's hosed
  162. SPEC!! D Non-clique topologies
  163. D Implement our own memory management, at least for common structs
  164. (Not ever necessary?)
  165. D Advanced directory servers
  166. D Automated reputation management
  167. SPEC!! D Figure out how to do threshold directory servers
  168. D jurisdiction info in dirserver entries? other info?
  169. Older (done) todo stuff:
  170. For 0.0.2pre17:
  171. o Put a H(K | handshake) into the onionskin response
  172. o Make cells 512 bytes
  173. o Reduce streamid footprint from 7 bytes to 2 bytes
  174. X Check for collisions in streamid (now possible with
  175. just 2 bytes), and back up & replace with padding if so
  176. o Use the 4 reserved bytes in each cell header to keep 1/5
  177. of a sha1 of the ongoing relay payload (move into stream header)
  178. o Move length into the stream header too
  179. o Make length 2 bytes
  180. D increase DH key length
  181. D increase RSA key length
  182. D Spec the stream_id stuff. Clarify that nobody on the backward
  183. stream should look at stream_id.
  184. Cell:
  185. ACI (anonymous circuit identifier) [2 bytes]
  186. Command [1 byte]
  187. Payload (padded with 0 bytes) [509 bytes]
  188. Relay payload:
  189. Relay command [1 byte]
  190. Stream ID [7 bytes]
  191. Partial SHA-1 [4 bytes]
  192. Length [2 bytes]
  193. Relay payload [495 bytes]
  194. For 0.0.2pre15:
  195. o don't pick exit nodes which will certainly reject all things.
  196. o don't pick nodes that the directory says are down
  197. o choose randomly from running dirservers, not just first one
  198. o install the man page
  199. o warn when client-side tries an address/port which no router in the dir accepts.
  200. For 0.0.2pre14:
  201. o More flexible exit policies (18.*, 18.0.0.0/8)
  202. o Work to succeed in the precense of exit policy violation
  203. o Replace desired_path_len with opaque path-selection specifier
  204. o Client-side DNS caching
  205. o Add entries to client DNS cache based on END cells
  206. o Remove port from END_REASON_EXITPOLICY cells
  207. o Start building new circuits when we get an exit-policy
  208. failure. (Defer exiting from the middle of existing
  209. circuits or extending existing circuits for later.)
  210. o Implement function to check whether a routerinfo_t
  211. supports a given exit addr.
  212. o Choose the exit node of an in-progress circuit based on
  213. pending AP connections.
  214. o Choose the exit node _first_, then beginning, then
  215. middle nodes.
  216. Previous:
  217. o Get tor to act like a socks server
  218. o socks4, socks4a
  219. o socks5
  220. o routers have identity key, link key, onion key.
  221. o link key certs are
  222. D signed by identity key
  223. D not in descriptor
  224. o not in config
  225. D not on disk
  226. o identity and onion keys are in descriptor (and disk)
  227. o upon boot, if it doesn't find identity key, generate it and write it.
  228. o also write a file with the identity key fingerprint in it
  229. o router generates descriptor: flesh out router_get_my_descriptor()
  230. o Routers sign descriptors with identity key
  231. o routers put version number in descriptor
  232. o routers should maybe have `uname -a` in descriptor?
  233. o Give nicknames to routers
  234. o in config
  235. o in descriptors
  236. o router posts descriptor
  237. o when it boots
  238. o every DirFetchPostPeriod seconds
  239. D when it changes
  240. o change tls stuff so certs don't get written to disk, or read from disk
  241. o make directory.c 'thread'safe
  242. o dirserver parses descriptor
  243. o dirserver checks signature
  244. D client checks signature?
  245. o dirserver writes directory to file
  246. o reads that file upon boot
  247. o directory includes all routers, up and down
  248. o add "up" line to directory, listing nicknames
  249. o instruments ORs to report stats
  250. o average cell fullness
  251. o average bandwidth used
  252. o configure log files. separate log file, separate severities.
  253. o what assumptions break if we fclose(0) when we daemonize?
  254. o make buffer struct elements opaque outside buffers.c
  255. o add log convention to the HACKING file
  256. o make 'make install' do the right thing
  257. o change binary name to tor
  258. o change config files so you look at commandline, else look in
  259. /etc/torrc. no cascading.
  260. o have an absolute datadir with fixed names for files, and fixed-name
  261. keydir under that with fixed names
  262. o Move (most of) the router/directory code out of main.c
  263. o Simple directory servers
  264. o Include key in source; sign directories
  265. o Signed directory backend
  266. o Document
  267. o Integrate
  268. o Add versions to code
  269. o Have directories list recommended-versions
  270. o Include line in directories
  271. o Check for presence of line.
  272. o Quit if running the wrong version
  273. o Command-line option to override quit
  274. o Add more information to directory server entries
  275. o Exit policies
  276. o Clearer bandwidth management
  277. o Do we want to remove bandwidth from OR handshakes?
  278. o What about OP handshakes?
  279. X Move away from openssl
  280. o Abstract out crypto calls
  281. X Look at nss, others? Just include code?
  282. o Use a stronger cipher
  283. o aes now, by including the code ourselves
  284. X On the fly compression of each stream
  285. o Clean up the event loop (optimize and sanitize)
  286. o Remove that awful concept of 'roles'
  287. o Terminology
  288. o Circuits, topics, cells stay named that
  289. o 'Connection' gets divided, or renamed, or something?
  290. o DNS farm
  291. o Distribute queries onto the farm, get answers
  292. o Preemptively grow a new worker before he's needed
  293. o Prune workers when too many are idle
  294. o DNS cache
  295. o Clear DNS cache over time
  296. D Honor DNS TTL info (how??)
  297. o Have strategy when all workers are busy
  298. o Keep track of which connections are in dns_wait
  299. o Need to cache positives/negatives on the tor side
  300. o Keep track of which queries have been asked
  301. o Better error handling when
  302. o An address doesn't resolve
  303. o We have max workers running
  304. o Consider taking the master out of the loop?
  305. X Implement reply onions
  306. o Total rate limiting
  307. o Look at OR handshake in more detail
  308. o Spec it
  309. o Merge OR and OP handshakes
  310. o rearrange connection_or so it doesn't suck so much to read
  311. D Periodic link key rotation. Spec?
  312. o wrap malloc with something that explodes when it fails
  313. o Clean up the number of places that get to look at prkey