TODO 20 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414
  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. Flag-day changes: (things which are backward incompatible)
  13. o remove link key from directories, from connection_t.
  14. (just get it from the tls cert)
  15. o Generate link keys on startup; don't store them to disk.
  16. o make onion keys include oaep padding, so you can tell
  17. if you decrypted it correctly
  18. o Rotate onion keys as needed
  19. - Rotate TLS connections [arma]
  20. - Set expiration times on X509 certs [nickm]
  21. . add bandwidthrate and bandwidthburst to server descriptor [nickm]
  22. - directories need to say who signed them. [nickm]
  23. - what other pieces of the descriptors need to change?
  24. maybe add a section for who's connected to a given router?
  25. add a flexible section for reputation info?
  26. Bugs:
  27. o we call signal(), but we should be calling sigaction()
  28. o send socks rejects when things go bad ?
  29. o on solaris, need to build with
  30. LDFLAGS="-lsocket -lnsl" ./configure
  31. o on solaris, we HAVE_UNAME but the uname() call fails?
  32. . should maybe make clients exit(1) when bad things happen?
  33. e.g. clock skew.
  34. o client-side dns cache doesn't appear to be getting populated
  35. by 'connected' cells. In fact, the 'connected' cells don't even
  36. include the IP.
  37. o When it can't resolve any dirservers, it is useless from then on.
  38. We should make it reload the RouterFile if it has no dirservers.
  39. o Sometimes it picks a middleman node as the exit for a circuit.
  40. o if you specify a non-dirserver as exitnode or entrynode, when it
  41. makes the first few circuits it hasn't yet fetched the directory,
  42. so it warns that it doesn't know the node.
  43. o make 'make test' exit(1) if a test fails.
  44. . fix buffer unit test so it passes
  45. Short-term:
  46. - should retry exitpolicy end streams even if the end cell didn't
  47. resolve the address for you
  48. o add in 'notice' log level
  49. X make recommendedversions different for clients and servers.
  50. e.g. C0.0.3 vs S0.0.3?
  51. o put IP into the descriptor, so clients don't need to resolve things
  52. o when you hup, rewrite the router.desc file (and maybe others)
  53. - consider handling broken socks4 implementations
  54. o improve how it behaves when i remove a line from the approved-routers files
  55. - Make tls connections tls_close intentionally
  56. - Add '[...truncated]' or similar to truncated log entries (like the directory
  57. in connection_dir_process_inbuf()).
  58. . Make logs handle it better when writing to them fails.
  59. o leave server descriptor out of directory if it's too old
  60. o Rename ACI to circID
  61. o integrate rep_ok functions, see what breaks
  62. - update tor faq
  63. o obey SocksBindAddress, ORBindAddress
  64. o warn if we're running as root
  65. o make connection_flush_buf() more obviously obsolete
  66. o let hup reread the config file, eg so we can get new exit
  67. policies without restarting
  68. o Put recommended_versions in a config entry
  69. X use times(2) rather than gettimeofday to measure how long it
  70. takes to process a cell
  71. o Separate trying to rebuild a circuit because you have none from trying
  72. to rebuild a circuit because the current one is stale
  73. X Continue reading from socks port even while waiting for connect.
  74. o Exit policies
  75. o Spec how to write the exit policies
  76. o Path selection algorithms
  77. o Choose path more incrementally
  78. o Let user request first/last node
  79. o And disallow certain nodes
  80. D Choose path by jurisdiction, etc?
  81. o Make relay end cells have failure status and payload attached
  82. X let non-approved routers handshake.
  83. X Dirserver shouldn't put you in running-routers list if you haven't
  84. uploaded a descriptor recently
  85. X migrate to using nickname rather than addr:port for routers
  86. - migrate to using IPv6 sizes everywhere
  87. o Move from onions to ephemeral DH
  88. o incremental path building
  89. o transition circuit-level sendmes to hop-level sendmes
  90. o implement truncate, truncated
  91. o move from 192byte DH to 128byte DH, so it isn't so damn slow
  92. X exiting from not-last hop
  93. X OP logic to decide to extend/truncate a path
  94. X make sure exiting from the not-last hop works
  95. X logic to find last *open* hop, not last hop, in cpath
  96. o Remember address and port when beginning.
  97. - Extend by nickname/hostname/something, not by IP.
  98. - Need a relay teardown cell, separate from one-way ends.
  99. X remove per-connection rate limiting
  100. - Make it harder to circumvent bandwidth caps: look at number of bytes
  101. sent across sockets, not number sent inside TLS stream.
  102. o Audit users of connnection_remove and connection_free outside of
  103. main.c; many should use mark_for_close instead.
  104. Rendezvous service:
  105. o Design and specify protocol
  106. o Possible preliminary refactoring:
  107. o Should we break circuits up into "circuit-with-cpath" and
  108. "circuit-without-cpath"?
  109. o We need a way to tag circuits as special-purpose circuits for:
  110. o Connecting from Bob's OP to the introduction point
  111. o Sending introduction requests from the IPoint to Bob
  112. o Connecting from Alice to the rendezvous point for Bob
  113. o Connecting from Bob to the rendezvous point for Alice
  114. o Waiting at a rendezvous point to be joined
  115. o Joined to another circuit at the rendezvous point.
  116. (We should also enumerate all the states that these operations
  117. can be in.) [NM]
  118. o Add circuit metadata [NM]
  119. o Code to configure hidden services [NM] 4 hours
  120. o Service descriptors
  121. o OPs need to maintain identity keys for hidden services [NM]
  122. o Code to generate and parse service descriptors [NM]
  123. o Advertisement
  124. o Generate y.onion hostnames [NM]
  125. o Store y.onion hostnames to disk. [NM]
  126. o Code to do an HTTP connection over Tor from within Tor [RD]
  127. o Publish service descriptors to directory [RD]
  128. o Directory accepts and remembers service descriptors, and
  129. delivers them as requested
  130. o Frontend [RD]
  131. o Backend [NM]
  132. o Code for OPs to retrieve (and cache?) service descriptors [RD]
  133. o Rendezvous
  134. o Code as needed to generate and parse all rendezvous-related
  135. cell types, and do all handshaking [NM]
  136. o ORs implement introduction points
  137. o OPs with hidden services establish introduction points
  138. o ORs implement rendezvous points
  139. o OPs notice y.onion URLs, and:
  140. o Retrieve service descriptors
  141. o Establish rendezvous points
  142. o Send introduction requests to introduction points
  143. o Communication
  144. o OPs remember which circuits are used for which rendezvous
  145. points, and can look up circuits by location-hidden service
  146. o OPs send/handle BEGIN cells for location-hidden services
  147. o End-to-end communication for location-hidden services
  148. o a section in the man pages: how to configure hidden services
  149. o let bob use himself as a rendezvous point
  150. o let bob choose himself as intro point
  151. o let bob replenish his intro points and republish
  152. o alice retries introduction and rendezvous a few times?
  153. o ORs should not pick themselves while building general circs
  154. o should alice ever try to refresh her service desc cache entries?
  155. should she expire them after e.g. 15 mins?
  156. o race condition: alice has the serverdesc in her cache, she opens
  157. the circs, serverdesc expires and is flushed, then she goes
  158. to send the intro cell. should serverdesc cache have a
  159. last-touched field? are there better fixes?
  160. o backward compatibility: when only certain nodes know about rend
  161. protocol, how do we deal? have nodes parse the tor version field?
  162. force an upgrade? simply be more robust against useless nodes?
  163. o should expire rend streams when too much time has passed
  164. o should make failed rend/intro circs count toward alice's
  165. num_failed circs, to prevent madness when we're offline (But
  166. don't count failed rend circs toward Bob's total, or Alice
  167. can bork him.)
  168. o deal with edge_type in connection_edge.c
  169. o retry end for certain reasons (resolvefailed, policyfailed)
  170. - preemptively build and start rendezvous circs
  171. - preemptively build n-1 hops of intro circs?
  172. o (n)ack introduction requests?
  173. - cannibalize general circs?
  174. D how to set up multiple locations for a hidden service?
  175. o make bob publish only established intro circs?
  176. o when bob tries to connect to alice's chosen rend point, but
  177. can't, but it's not the fault of the last hop in the rend
  178. circ, then he should retry?
  179. - fix router_get_by_* functions so they can get ourselves too,
  180. and audit everything to make sure rend and intro points are
  181. just as likely to be us as not.
  182. On-going
  183. . Better comments for functions!
  184. . Go through log messages, reduce confusing error messages.
  185. . make the logs include more info (fd, etc)
  186. . Unit tests
  187. . Update the spec so it matches the code
  188. Mid-term:
  189. o Refactor: add own routerinfo to routerlist. Right now, only
  190. router_get_by_nickname knows about 'this router', as a hack to
  191. get circuit_launch_new to do the right thing.
  192. - Rotate tls-level connections -- make new ones, expire old ones.
  193. So we get actual key rotation, not just symmetric key rotation
  194. - And learn to transfer a circuit from one conn to another, so we
  195. can empty conns to expire them.
  196. o Are there anonymity issues with sequential streamIDs? Sequential
  197. circIDs? Eg an attacker can learn how many there have been.
  198. The fix is to initialize them randomly rather than at 1.
  199. - Look at having smallcells and largecells
  200. . Redo scheduler
  201. o fix SSL_read bug for buffered records
  202. - make round-robining more fair
  203. - What happens when a circuit's length is 1? What breaks?
  204. . streams / circuits
  205. o Implement streams
  206. o Rotate circuits after N minutes?
  207. X Circuits should expire when circuit->expire triggers
  208. NICK . Handle half-open connections
  209. o openssh is an application that uses half-open connections
  210. o Figure out what causes connections to close, standardize
  211. when we mark a connection vs when we tear it down
  212. o Look at what ssl does to keep from mutating data streams
  213. o Put CPU workers in separate processes
  214. o Handle multiple cpu workers (one for each cpu, plus one)
  215. o Queue for pending tasks if all workers full
  216. o Support the 'process this onion' task
  217. D Merge dnsworkers and cpuworkers to some extent
  218. o Handle cpuworkers dying
  219. . Scrubbing proxies
  220. - Find an smtp proxy?
  221. - Check the old smtp proxy code
  222. o Find an ftp proxy? wget --passive
  223. D Wait until there are packet redirectors for Linux
  224. . Get socks4a support into Mozilla
  225. . Tests
  226. o Testing harness/infrastructure
  227. D System tests (how?)
  228. - Performance tests, so we know when we've improved
  229. . webload infrastructure (Bruce)
  230. . httperf infrastructure (easy to set up)
  231. . oprofile (installed in RH >8.0)
  232. NICK . Daemonize and package
  233. o Teach it to fork and background
  234. . Red Hat spec file
  235. o Debian spec file equivalent
  236. . Portability
  237. . Which .h files are we actually using?
  238. . Port to:
  239. o Linux
  240. o BSD
  241. o Solaris
  242. o Cygwin
  243. . Win32
  244. o OS X
  245. - deal with pollhup / reached_eof on all platforms
  246. o openssl randomness
  247. o inet_ntoa
  248. o stdint.h
  249. - Make a script to set up a local network on your machine
  250. o More flexibility in node addressing
  251. D Support IPv6 rather than just 4
  252. o Handle multihomed servers (config variable to set IP)
  253. In the distant future:
  254. D tunnel tor cell protocol over http, for people who need to
  255. do http
  256. D better transport than tcp: reliable is necessary, but
  257. out-of-order delivery is fine (to some extent).
  258. D Load balancing between router twins
  259. D Keep track of load over links/nodes, to
  260. know who's hosed
  261. SPEC!! D Non-clique topologies
  262. D Advanced directory servers
  263. D Automated reputation management
  264. SPEC!! D Figure out how to do threshold directory servers
  265. D jurisdiction info in dirserver entries? other info?
  266. Older (done) todo stuff:
  267. For 0.0.2pre17:
  268. o Put a H(K | handshake) into the onionskin response
  269. o Make cells 512 bytes
  270. o Reduce streamid footprint from 7 bytes to 2 bytes
  271. X Check for collisions in streamid (now possible with
  272. just 2 bytes), and back up & replace with padding if so
  273. o Use the 4 reserved bytes in each cell header to keep 1/5
  274. of a sha1 of the ongoing relay payload (move into stream header)
  275. o Move length into the stream header too
  276. o Make length 2 bytes
  277. D increase DH key length
  278. D increase RSA key length
  279. D Spec the stream_id stuff. Clarify that nobody on the backward
  280. stream should look at stream_id.
  281. For 0.0.2pre15:
  282. o don't pick exit nodes which will certainly reject all things.
  283. o don't pick nodes that the directory says are down
  284. o choose randomly from running dirservers, not just first one
  285. o install the man page
  286. o warn when client-side tries an address/port which no router in the dir accepts.
  287. For 0.0.2pre14:
  288. o More flexible exit policies (18.*, 18.0.0.0/8)
  289. o Work to succeed in the precense of exit policy violation
  290. o Replace desired_path_len with opaque path-selection specifier
  291. o Client-side DNS caching
  292. o Add entries to client DNS cache based on END cells
  293. o Remove port from END_REASON_EXITPOLICY cells
  294. o Start building new circuits when we get an exit-policy
  295. failure. (Defer exiting from the middle of existing
  296. circuits or extending existing circuits for later.)
  297. o Implement function to check whether a routerinfo_t
  298. supports a given exit addr.
  299. o Choose the exit node of an in-progress circuit based on
  300. pending AP connections.
  301. o Choose the exit node _first_, then beginning, then
  302. middle nodes.
  303. Previous:
  304. o Get tor to act like a socks server
  305. o socks4, socks4a
  306. o socks5
  307. o routers have identity key, link key, onion key.
  308. o link key certs are
  309. D signed by identity key
  310. D not in descriptor
  311. o not in config
  312. D not on disk
  313. o identity and onion keys are in descriptor (and disk)
  314. o upon boot, if it doesn't find identity key, generate it and write it.
  315. o also write a file with the identity key fingerprint in it
  316. o router generates descriptor: flesh out router_get_my_descriptor()
  317. o Routers sign descriptors with identity key
  318. o routers put version number in descriptor
  319. o routers should maybe have `uname -a` in descriptor?
  320. o Give nicknames to routers
  321. o in config
  322. o in descriptors
  323. o router posts descriptor
  324. o when it boots
  325. o every DirFetchPostPeriod seconds
  326. D when it changes
  327. o change tls stuff so certs don't get written to disk, or read from disk
  328. o make directory.c 'thread'safe
  329. o dirserver parses descriptor
  330. o dirserver checks signature
  331. D client checks signature?
  332. o dirserver writes directory to file
  333. o reads that file upon boot
  334. o directory includes all routers, up and down
  335. o add "up" line to directory, listing nicknames
  336. o instruments ORs to report stats
  337. o average cell fullness
  338. o average bandwidth used
  339. o configure log files. separate log file, separate severities.
  340. o what assumptions break if we fclose(0) when we daemonize?
  341. o make buffer struct elements opaque outside buffers.c
  342. o add log convention to the HACKING file
  343. o make 'make install' do the right thing
  344. o change binary name to tor
  345. o change config files so you look at commandline, else look in
  346. /etc/torrc. no cascading.
  347. o have an absolute datadir with fixed names for files, and fixed-name
  348. keydir under that with fixed names
  349. o Move (most of) the router/directory code out of main.c
  350. o Simple directory servers
  351. o Include key in source; sign directories
  352. o Signed directory backend
  353. o Document
  354. o Integrate
  355. o Add versions to code
  356. o Have directories list recommended-versions
  357. o Include line in directories
  358. o Check for presence of line.
  359. o Quit if running the wrong version
  360. o Command-line option to override quit
  361. o Add more information to directory server entries
  362. o Exit policies
  363. o Clearer bandwidth management
  364. o Do we want to remove bandwidth from OR handshakes?
  365. o What about OP handshakes?
  366. X Move away from openssl
  367. o Abstract out crypto calls
  368. X Look at nss, others? Just include code?
  369. o Use a stronger cipher
  370. o aes now, by including the code ourselves
  371. X On the fly compression of each stream
  372. o Clean up the event loop (optimize and sanitize)
  373. o Remove that awful concept of 'roles'
  374. o Terminology
  375. o Circuits, topics, cells stay named that
  376. o 'Connection' gets divided, or renamed, or something?
  377. o DNS farm
  378. o Distribute queries onto the farm, get answers
  379. o Preemptively grow a new worker before he's needed
  380. o Prune workers when too many are idle
  381. o DNS cache
  382. o Clear DNS cache over time
  383. D Honor DNS TTL info (how??)
  384. o Have strategy when all workers are busy
  385. o Keep track of which connections are in dns_wait
  386. o Need to cache positives/negatives on the tor side
  387. o Keep track of which queries have been asked
  388. o Better error handling when
  389. o An address doesn't resolve
  390. o We have max workers running
  391. o Consider taking the master out of the loop?
  392. X Implement reply onions
  393. o Total rate limiting
  394. o Look at OR handshake in more detail
  395. o Spec it
  396. o Merge OR and OP handshakes
  397. o rearrange connection_or so it doesn't suck so much to read
  398. D Periodic link key rotation. Spec?
  399. o wrap malloc with something that explodes when it fails
  400. o Clean up the number of places that get to look at prkey