TODO 24 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521
  1. $Id$
  2. Legend:
  3. SPEC!! - Not specified
  4. SPEC - Spec not finalized
  5. N - nick claims
  6. R - arma claims
  7. P - phobos claims
  8. - Not done
  9. * Top priority
  10. . Partially done
  11. o Done
  12. d Deferrable
  13. D Deferred
  14. X Abandoned
  15. Items for 0.1.2.x-rc:
  16. ? - Bug: combination of things:
  17. When we've been idle a long time, we stop fetching server
  18. descriptors. When we then get a socks request, we build circuits
  19. immediately using whatever descriptors we have, rather than waiting
  20. until we've fetched correct ones.
  21. N - Test guard unreachable logic; make sure that we actually attempt to
  22. connect to guards that we think are unreachable from time to time.
  23. Make sure that we don't freak out when the network is down.
  24. o weight dir requests by advertised bandwidth? with maybe a lower cutoff
  25. than for tor traffic. perhaps also weighted by the expected size of
  26. the response.
  27. . Have (and document) a BEGIN_DIR relay cell that means "Connect to your
  28. directory port."
  29. o Implement
  30. R - turn the received socks addr:port into a digest for setting .exit
  31. - be able to connect without having a server descriptor, to bootstrap.
  32. R - handle connect-dir streams that don't have a chosen_exit_name set.
  33. o include ORPort in DirServers lines so we can know where to connect.
  34. list the orport as 0 if it can't handle begin_dir.
  35. o List orports of actual dirservers..
  36. . option to dl directory info via tor:
  37. TunnelDirConns and PreferTunneledDirConns
  38. R - actually cause the directory.c functions to know about or_port
  39. and use it when we're supposed to.
  40. o for tunneled edge conns, stop reading to the bridge connection
  41. when the or_conn we're writing to has a full outbuf.
  42. o make directory bridge data not get produced when the corresponding
  43. or_conn is full, and accept the sometimes directory data will just
  44. never get written.
  45. o DNS improvements
  46. o Asynchronous DNS
  47. X Make evdns use windows strerror equivalents.
  48. (there don't seem to be windows equivalents for strerror on the
  49. WSAERR* codes.)
  50. o Make sure patches get into libevent.
  51. o Verify that it works well on windows
  52. o Debug and re-enable server-side reverse DNS caching
  53. - Critical but minor bugs, backport candidates.
  54. - support dir 503s better
  55. o clients don't log as loudly when they receive them
  56. o they don't count toward the 3-strikes rule
  57. D But eventually, we give up after getting a lot of 503s.
  58. D Delay when we get a lot of 503s, rather than punting onto the
  59. servers that have given us 503s?
  60. o Add a 'BadDirectory' flag to statuses.
  61. o authorities should *never* 503 a cache, and should never 503
  62. network status requests.
  63. D They can 503 client descriptor requests when they feel like it.
  64. How can they distinguish? Not implemented for now, maybe
  65. should abandon.
  66. - update dir-spec with what we decided for each of these
  67. o Controller fixes
  68. o Make REMAP stream events have a SOURCE (cache or exit), and make them
  69. generated in every case where we get a connected or resolved cell.
  70. N - Libevent fixes
  71. . Try to get the pthread_sigprocmask situation under control; see
  72. coderman's or-dev post of 20-Feb-2007.
  73. - Polishing
  74. - Profile client and server; fix slow spots
  75. - Address XXX012 items
  76. . Packaging
  77. o Switch canonical win32 compiler to mingw.
  78. - If we haven't replaced privoxy, lock down its configuration in all
  79. packages, as documented in tor-doc-unix.html
  80. . Forward compatibility fixes
  81. - Start uploading short and long descriptors; authorities should support
  82. URLs to retrieve long descriptors, and should discard short descriptors
  83. for now. Later, once tools use the "long descriptor" URLs, authorities
  84. will serve the short descriptors every time they're asked for
  85. a descriptor.
  86. NR - Design
  87. N - Implement, if we think it's smart.
  88. o Check for any outstanding checks we do on the form or number of client
  89. certificates that would prevent us from executing certain
  90. blocking-resistance strategies.
  91. o Design (proposal 106)
  92. o Implement
  93. N - Hack up a client that gives out weird/no certificates, so we can
  94. test to make sure that this doesn't cause servers to crash.
  95. For 0.1.2.x-final:
  96. NR. Write path-spec.txt
  97. - Docs
  98. - Tell people about OSX Uninstaller
  99. - Quietly document NT Service options
  100. - More prominently, we should have a recommended apps list.
  101. - recommend gaim.
  102. - unrecommend IE because of ftp:// bug.
  103. N - we should add a preamble to tor-design saying it's out of date.
  104. N . Document transport and natdport
  105. o In man page
  106. - In a good HOWTO.
  107. Things we'd like to do in 0.2.0.x:
  108. - Proposals:
  109. - 101: Voting on the Tor Directory System
  110. - 104: Long and Short Router Descriptors
  111. - 105: Version negotiation for the Tor protocol
  112. - Refactoring:
  113. - Make resolves no longer use edge_connection_t unless needed.
  114. - Make cells get buffered on circuit, not on the or_conn.
  115. - Move all status info out of routerinfo into local_routerstatus. Make
  116. "who can change what" in local_routerstatus explicit. Make
  117. local_routerstatus (or equivalent) subsume all places to go for "what
  118. router is this?"
  119. - Remove socketpair-based bridges conns, and the word "bridge".
  120. - Generate torrc.{complete|sample}.in, tor.1.in, the HTML manual, and the
  121. online config documentation from a single source.
  122. - Features:
  123. - Implement a DNS proxy
  124. - Bridges.
  125. - Let controller set router flags for authority to transmit, and for
  126. client to use.
  127. - Support relaying streams to ipv6.
  128. - Deprecations:
  129. - Remove v0 control protocol.
  130. Deferred from 0.1.2.x:
  131. - Improve autoconf process to handle multiple SSL installations better.
  132. X <nickm> "Let's try to find a way to make it run and make the version
  133. match, but if not, let's just make it run."
  134. X <arma> "should we detect if we have a --with-ssl-dir and try the -R
  135. by default, if it works?"
  136. - finish status event implementation and accompanying getinfos
  137. - More work on AvoidDiskWrites?
  138. - Get some kind of "meta signing key" to be used solely to sign
  139. releases/to certify releases when signed by the right people/
  140. to certify sign the right people's keys? Also use this to cert the SSL
  141. key, etc. (Proposal 103)
  142. - per-conn write buckets
  143. - separate config options for read vs write limiting
  144. (It's hard to support read > write, since we need better
  145. congestion control to avoid overfull buffers there. So,
  146. defer the whole thing.)
  147. P - Figure out why dll's compiled in mingw don't work right in WinXP.
  148. P - Figure out why openssl 0.9.8d "make test" fails at sha256t test.
  149. - don't do dns hijacking tests if we're reject *:* exit policy?
  150. (deferred until 0.1.1.x is less common)
  151. - Directory guards
  152. - RAM use in directory authorities.
  153. - Memory use improvements:
  154. - Look into pulling serverdescs off buffers as they arrive.
  155. - Save and mmap v1 directories, and networkstatus docs; store them
  156. zipped, not uncompressed.
  157. - Switch cached_router_t to use mmap.
  158. - What to do about reference counts on windows? (On Unix, this is
  159. easy: unlink works fine. (Right?) On Windows, I have doubts. Do we
  160. need to keep multiple files?)
  161. - What do we do about the fact that people can't read zlib-
  162. compressed files manually?
  163. - Change the way we handle cells, flow-control, and bridges.
  164. - The issue is that we package from edge connections aggressively until
  165. we hit their package windows or the circuit package windows, even if
  166. the buffer on the corresponding OR connection is pretty damn big. This
  167. sucks from a RAM usage POV. Now, we could try to stop reading on the
  168. edges (or just the edges connected to a local bridge) when an or_conn's
  169. outbuf is full. But if we're a server, and we stop reading on some
  170. exit conns when OR conns are full, soon OR conns will contain only
  171. traffic from other OR conns, and the exit data in question will never
  172. get written.
  173. - Remove socketpair-based bridges: use shared (or connected) buffers for
  174. communication, rather than sockets.
  175. - When relaying cells from an OR conn to an OR conn, have them wait in a
  176. queue on the or_circuit_t object; don't move them onto the target conn
  177. until there is space in the target conn's outbuf. Also, only package
  178. data from exitconns when there is space in the target conn's outbuf.
  179. - As an added advantage, this would let us kill stalled _circuits_
  180. when their buffers get too full, rather than killing entire OR
  181. conns. But we must think about anonymity implications of that.
  182. - We'll probably want to do some kind of big refactoring of our
  183. dataflow when we do these changes; stuff is hairy enough already,
  184. and it will only get harrier with this stuff.
  185. - If the client's clock is too far in the past, it will drop (or
  186. just not try to get) descriptors, so it'll never build circuits.
  187. - Tolerate clock skew on bridge relays.
  188. - A more efficient dir protocol. (Proposal 101)
  189. - Authorities should fetch the network-statuses amongst each
  190. other, consensus them, and advertise a communal network-status.
  191. This is not so much for safety/complexity as it is to reduce
  192. bandwidth requirements for Alice.
  193. - How does this interact with our goal of being able to choose
  194. your own dir authorities? I guess we're now assuming that all
  195. dir authorities know all the other authorities in their "group"?
  196. - Should we also look into a "delta since last network-status
  197. checkpoint" scheme, to reduce overhead further?
  198. - Extend the "r" line in network-status to give a set of buckets (say,
  199. comma-separated) for that router.
  200. - Buckets are deterministic based on IP address.
  201. - Then clients can choose a bucket (or set of buckets) to
  202. download and use.
  203. - Improvements to versioning. (Proposal 105)
  204. - When we connect to a Tor server, it sends back a cell listing
  205. the IP it believes it is using. Use this to block dvorak's attack.
  206. Also, this is a fine time to say what time you think it is.
  207. o Verify that a new cell type is okay with deployed codebase
  208. . Specify HELLO cells
  209. . Figure out v0 compatibility.
  210. - Implement
  211. - Eventdns improvements
  212. - Have a way to query for AAAA and A records simultaneously.
  213. - Improve request API: At the very least, add the ability to construct
  214. a more-or-less arbitrary request and get a response.
  215. - (Can we suppress cnames? Should we?)
  216. - Now that we're avoiding exits when picking non-exit positions,
  217. we need to consider how to pick nodes for internal circuits. If
  218. we avoid exits for all positions, we skew the load balancing. If
  219. we accept exits for all positions, we leak whether it's an internal
  220. circuit at every step. If we accept exits only at the last hop, we
  221. reintroduce Lasse's attacks from the Oakland paper.
  222. - We should ship with a list of stable dir mirrors -- they're not
  223. trusted like the authorities, but they'll provide more robustness
  224. and diversity for bootstrapping clients.
  225. - Simplify authority operation
  226. - Follow weasel's proposal, crossed with mixminion dir config format
  227. - A way to adjust router flags from the controller.
  228. (How do we prevent the authority from clobbering them soon after?)
  229. - a way to pick entry guards based wholly on extend_info equivalent;
  230. a way to export extend_info equivalent.
  231. - Better estimates in the directory of whether servers have good uptime
  232. (high expected time to failure) or good guard qualities (high
  233. fractional uptime).
  234. - AKA Track uptime as %-of-time-up, as well as time-since-last-down
  235. - Have a "Faster" status flag that means it. Fast2, Fast4, Fast8?
  236. - spec
  237. - implement
  238. - Failed rend desc fetches sometimes don't get retried. True/false?
  239. - Windows server usability
  240. - Solve the ENOBUFS problem.
  241. - make tor's use of openssl operate on buffers rather than sockets,
  242. so we can make use of libevent's buffer paradigm once it has one.
  243. - make tor's use of libevent tolerate either the socket or the
  244. buffer paradigm; includes unifying the functions in connect.c.
  245. - We need a getrlimit equivalent on Windows so we can reserve some
  246. file descriptors for saving files, etc. Otherwise we'll trigger
  247. asserts when we're out of file descriptors and crash.
  248. M - rewrite how libevent does select() on win32 so it's not so very slow.
  249. - Add overlapped IO
  250. - Add an option (related to AvoidDiskWrites) to disable directory caching.
  251. - More status event features:
  252. - Missing events:
  253. - DIR_REACHABLE
  254. - BAD_DIR_RESPONSE (Unexpected directory response; maybe we're behind
  255. a firewall.)
  256. - BAD_PROXY (Bad http or https proxy)
  257. - UNRECOGNIZED_ROUTER (a nickname we asked for is unavailable)
  258. - Status events related to hibernation
  259. - something about failing to parse our address?
  260. from resolve_my_address() in config.c
  261. - sketchy OS, sketchy threading
  262. - too many onions queued: threading problems or slow CPU?
  263. - Missing fields:
  264. - TIMEOUT on CHECKING_REACHABILITY
  265. - GETINFO status/client, status/server, status/general: There should be
  266. some way to learn which status events are currently "in effect."
  267. We should specify which these are, what format they appear in, and so
  268. on.
  269. - Improvements to bandwidth counting
  270. - look into "uncounting" bytes spent on local connections, so
  271. we can bandwidthrate but still have fast downloads.
  272. - "bandwidth classes", for incoming vs initiated-here conns,
  273. and to give dir conns lower priority.
  274. Minor items for 0.1.2.x as time permits:
  275. - include bandwidth breakdown by conn->type in BW events.
  276. - getinfo ns/name/moria2 doesn't include a "v" line, even when some
  277. network-statuses I have show it. I suppose the fix should go in
  278. networkstatus_getinfo_helper_single() in routerlist.c.
  279. - Unify autoconf search code for libevent and openssl. Make code
  280. suggest platform-appropriate "devel" / "dev" / whatever packages
  281. if we can link but we can't find the headers.
  282. - Recommend polipo? Please?
  283. - Make documentation realize that location of system configuration file
  284. will depend on location of system defaults, and isn't always /etc/torrc.
  285. - Review torrc.sample to make it more discursive.
  286. o when reporting clock skew (both to logs and to controller), if it's
  287. taken 126 seconds to read from the directory, our clock skew estimate
  288. is 126 seconds wrong. use conn->timestamp_create or _lastwritten
  289. for a closer estimate?
  290. - a way to generate the website diagrams from source, so we can
  291. translate them as utf-8 text rather than with gimp.
  292. R - add d64 and fp64 along-side d and fp so people can paste status
  293. entries into a url. since + is a valid base64 char, only allow one
  294. at a time. spec and then do.
  295. - When we export something from foo.c file for testing purposes only,
  296. make a foo_test.h file for test.c to include.
  297. - The Debian package now uses --verify-config when (re)starting,
  298. to distinguish configuration errors from other errors. Perhaps
  299. the RPM and other startup scripts should too?
  300. - add a "default.action" file to the tor/vidalia bundle so we can fix the
  301. https thing in the default configuration:
  302. http://wiki.noreply.org/noreply/TheOnionRouter/TorFAQ#PrivoxyWeirdSSLPort
  303. . Flesh out options_description array in src/or/config.c
  304. o Don't let 'newnym' be triggered more often than every n seconds.
  305. X If we try to publish as a nickname that's already claimed, should
  306. we append a number (or increment the number) and try again? This
  307. way people who read their logs can fix it as before, but people
  308. who don't read their logs will still offer Tor servers.
  309. - Fall back to unnamed; warn user; send controller event. ("When we
  310. notice a 'Rejected: There is already a named server with this nickname'
  311. message... or maybe instead when we see in the networkstatuses that
  312. somebody else is Named with the name we want: warn the user, send a
  313. STATUS_SERVER message, and fall back to unnamed.")
  314. ! - Tor should bind its ports before dropping privs, so users don't
  315. have to do the ipchains dance.
  316. - Rate limit exit connections to a given destination -- this helps
  317. us play nice with websites when Tor users want to crawl them; it
  318. also introduces DoS opportunities.
  319. - Streamline how we pick entry nodes: Make choose_random_entry() have
  320. less magic and less control logic.
  321. - Christian Grothoff's attack of infinite-length circuit.
  322. the solution is to have a separate 'extend-data' cell type
  323. which is used for the first N data cells, and only
  324. extend-data cells can be extend requests.
  325. - Specify, including thought about anonymity implications.
  326. - Display the reasons in 'destroy' and 'truncated' cells under some
  327. circumstances?
  328. - We need a way for the authorities to declare that nodes are
  329. in a family. Also, it kinda sucks that family declarations use O(N^2)
  330. space in the descriptors.
  331. - If the server is spewing complaints about raising your ulimit -n,
  332. we should add a note about this to the server descriptor so other
  333. people can notice too.
  334. - cpu fixes:
  335. - see if we should make use of truncate to retry
  336. X kill dns workers more slowly
  337. . Directory changes
  338. . Some back-out mechanism for auto-approval
  339. - a way of rolling back approvals to before a timestamp
  340. - Consider minion-like fingerprint file/log combination.
  341. - packaging and ui stuff:
  342. . multiple sample torrc files
  343. . figure out how to make nt service stuff work?
  344. . Document it.
  345. - Vet all pending installer patches
  346. - Win32 installer plus privoxy, sockscap/freecap, etc.
  347. - Vet win32 systray helper code
  348. - Improve controller
  349. - a NEWSTATUS event similar to NEWDESC.
  350. - change circuit status events to give more details, like purpose,
  351. whether they're internal, when they become dirty, when they become
  352. too dirty for further circuits, etc.
  353. - What do we want here, exactly?
  354. - Specify and implement it.
  355. - Change stream status events analogously.
  356. - What do we want here, exactly?
  357. - Specify and implement it.
  358. - Make other events "better".
  359. - Change stream status events analogously.
  360. - What do we want here, exactly?
  361. - Specify and implement it.
  362. - Make other events "better" analogously
  363. - What do we want here, exactly?
  364. - Specify and implement it.
  365. . Expose more information via getinfo:
  366. - import and export rendezvous descriptors
  367. - Review all static fields for additional candidates
  368. - Allow EXTENDCIRCUIT to unknown server.
  369. - We need some way to adjust server status, and to tell tor not to
  370. download directories/network-status, and a way to force a download.
  371. - Make everything work with hidden services
  372. - Directory system improvements
  373. - config option to publish what ports you listen on, beyond
  374. ORPort/DirPort. It should support ranges and bit prefixes (?) too.
  375. - Parse this.
  376. - Relay this in networkstatus.
  377. - Be a DNS proxy.
  378. - Need a way to request address lookups (and allocate a stream ID for
  379. them) without having a corresponding client socket.
  380. - Once this is done, it would be nice to have a way to request address
  381. lookups from the controller without using SOCKS.
  382. - Refactor exit/entry side of DNS resolve: we don't need a connection_t;
  383. we can have an edge_connection_t and (say) dns_request_t both extend an
  384. edge_stream_t, and have p_streams and n_streams both be linked lists
  385. of edge_stream_t.
  386. Future version:
  387. - Configuration format really wants sections.
  388. - Good RBL substitute.
  389. - Authorities should try using exits for http to connect to some URLS
  390. (specified in a configuration file, so as not to make the List Of Things
  391. Not To Censor completely obvious) and ask them for results. Exits that
  392. don't give good answers should have the BadExit flag set.
  393. - Our current approach to block attempts to use Tor as a single-hop proxy
  394. is pretty lame; we should get a better one.
  395. . Update the hidden service stuff for the new dir approach.
  396. - switch to an ascii format, maybe sexpr?
  397. - authdirservers publish blobs of them.
  398. - other authdirservers fetch these blobs.
  399. - hidserv people have the option of not uploading their blobs.
  400. - you can insert a blob via the controller.
  401. - and there's some amount of backwards compatibility.
  402. - teach clients, intro points, and hidservs about auth mechanisms.
  403. - come up with a few more auth mechanisms.
  404. - auth mechanisms to let hidden service midpoint and responder filter
  405. connection requests.
  406. - Bind to random port when making outgoing connections to Tor servers,
  407. to reduce remote sniping attacks.
  408. - Have new people be in limbo and need to demonstrate usefulness
  409. before we approve them.
  410. - Clients should estimate their skew as median of skew from servers
  411. over last N seconds.
  412. - Make router_is_general_exit() a bit smarter once we're sure what it's for.
  413. - Audit everything to make sure rend and intro points are just as likely to
  414. be us as not.
  415. - Do something to prevent spurious EXTEND cells from making middleman
  416. nodes connect all over. Rate-limit failed connections, perhaps?
  417. - Automatically determine what ports are reachable and start using
  418. those, if circuits aren't working and it's a pattern we recognize
  419. ("port 443 worked once and port 9001 keeps not working").
  420. - Limit to 2 dir, 2 OR, N SOCKS connections per IP.
  421. - Handle full buffers without totally borking
  422. - Rate-limit OR and directory connections overall and per-IP and
  423. maybe per subnet.
  424. - Hold-open-until-flushed now works by accident; it should work by
  425. design.
  426. - DoS protection: TLS puzzles, public key ops, bandwidth exhaustion.
  427. - Specify?
  428. - tor-resolve script should use socks5 to get better error messages.
  429. - hidserv offerers shouldn't need to define a SocksPort
  430. * figure out what breaks for this, and do it.
  431. - tor should be able to have a pool of outgoing IP addresses
  432. that it is able to rotate through. (maybe)
  433. - Specify; implement.
  434. - let each hidden service (or other thing) specify its own
  435. OutboundBindAddress?
  436. - Stop using tor_socketpair to make connection bridges: do an
  437. implementation that uses buffers only.
  438. Blue-sky:
  439. - Patch privoxy and socks protocol to pass strings to the browser.
  440. - Standby/hotswap/redundant hidden services.
  441. - Robust decentralized storage for hidden service descriptors.
  442. - The "China problem"
  443. - Allow small cells and large cells on the same network?
  444. - Cell buffering and resending. This will allow us to handle broken
  445. circuits as long as the endpoints don't break, plus will allow
  446. connection (tls session key) rotation.
  447. - Implement Morphmix, so we can compare its behavior, complexity, etc.
  448. - Other transport. HTTP, udp, rdp, airhook, etc. May have to do our own
  449. link crypto, unless we can bully openssl into it.
  450. - Need a relay teardown cell, separate from one-way ends.
  451. (Pending a user who needs this)
  452. - Handle half-open connections: right now we don't support all TCP
  453. streams, at least according to the protocol. But we handle all that
  454. we've seen in the wild.
  455. (Pending a user who needs this)
  456. Non-Coding:
  457. - Mark up spec; note unclear points about servers
  458. - Mention controller libs someplace.
  459. . more pictures from ren. he wants to describe the tor handshake
  460. NR- write a spec appendix for 'being nice with tor'
  461. - tor-in-the-media page
  462. - Remove need for HACKING file.
  463. - Figure out licenses for website material.
  464. - Specify the keys and key rotation schedules and stuff
  465. Website:
  466. - and remove home and make the "Tor" picture be the link to home.
  467. - put the logo on the website, in source form, so people can put it on
  468. stickers directly, etc.
  469. R - make a page with the hidden service diagrams.
  470. - ask Jan to be the translation coordinator? add to volunteer page.
  471. - add a page for localizing all tor's components.
  472. - It would be neat if we had a single place that described _all_ the
  473. tor-related tools you can use, and what they give you, and how well they
  474. work. Right now, we don't give a lot of guidance wrt
  475. torbutton/foxproxy/privoxy/polipo in any consistent place.