TODO 24 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514
  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. X <nickm> "Let's try to find a way to make it run and make the version
  16. match, but if not, let's just make it run."
  17. X <arma> "should we detect if we have a --with-ssl-dir and try the -R
  18. by default, if it works?"
  19. Items for 0.1.2.x, real soon now:
  20. ? - Bug: combination of things:
  21. When we've been idle a long time, we stop fetching server
  22. descriptors. When we then get a socks request, we build circuits
  23. immediately using whatever descriptors we have, rather than waiting
  24. until we've fetched correct ones.
  25. N - Test guard unreachable logic; make sure that we actually attempt to
  26. connect to guards that we think are unreachable from time to time.
  27. Make sure that we don't freak out when the network is down.
  28. Items for 0.1.2.x:
  29. o weight dir requests by advertised bandwidth? with maybe a lower cutoff
  30. than for tor traffic. perhaps also weighted by the expected size of
  31. the response.
  32. . Have (and document) a BEGIN_DIR relay cell that means "Connect to your
  33. directory port."
  34. o Implement
  35. R - turn the received socks addr:port into a digest for setting .exit
  36. - be able to connect without having a server descriptor, to bootstrap.
  37. R - handle connect-dir streams that don't have a chosen_exit_name set.
  38. o include ORPort in DirServers lines so we can know where to connect.
  39. list the orport as 0 if it can't handle begin_dir.
  40. o List orports of actual dirservers..
  41. o Servers are easy to setup and run: being a relay is about as easy as
  42. being a client.
  43. o Reduce resource load
  44. o A way to alert controller when router flags change.
  45. o Specify: SETEVENTS NS
  46. o Implement
  47. o Hunt for places that change networkstatus info that I might have
  48. missed.
  49. D "NS" appears to be used to notify of routerstatus updates.
  50. Do we want an actual NS event, for when new networkstatuses
  51. arrive, and then rename this one?
  52. (Renaming would be more trouble than it's worth, and a
  53. 'networkstatus arrived' event can wait till next series.
  54. . option to dl directory info via tor:
  55. TunnelDirConns and PreferTunneledDirConns
  56. R - actually cause the directory.c functions to know about or_port
  57. and use it when we're supposed to.
  58. o for tunneled edge conns, stop reading to the bridge connection
  59. when the or_conn we're writing to has a full outbuf.
  60. o make directory bridge data not get produced when the corresponding
  61. or_conn is full, and accept the sometimes directory data will just
  62. never get written.
  63. o When we get a connection using the v0 controller protocol, warn that the
  64. next revision of Tor won't support it.
  65. N - DNS improvements
  66. . Asynchronous DNS
  67. - Make evdns use windows strerror equivalents.
  68. o Make sure patches get into libevent.
  69. - Verify that it works well on windows
  70. o Debug and re-enable server-side reverse DNS caching
  71. - Critical but minor bugs, backport candidates.
  72. - support dir 503s better
  73. o clients don't log as loudly when they receive them
  74. o they don't count toward the 3-strikes rule
  75. D But eventually, we give up after getting a lot of 503s.
  76. D Delay when we get a lot of 503s, rather than punting onto the
  77. servers that have given us 503s?
  78. o Add a 'BadDirectory' flag to statuses.
  79. o authorities should *never* 503 a cache, and should never 503
  80. network status requests.
  81. D They can 503 client descriptor requests when they feel like it.
  82. How can they distinguish? Not implemented for now, maybe
  83. should abandon.
  84. - update dir-spec with what we decided for each of these
  85. NR. Write path-spec.txt
  86. - Polishing
  87. - Profile client and server; fix slow spots
  88. - Address XXX012 items
  89. - Packaging
  90. - Tell people about OSX Uninstaller
  91. - Quietly document NT Service options
  92. - Switch canonical win32 compiler to mingw.
  93. - If we haven't replaced privoxy, lock down its configuration in all
  94. packages, as documented in tor-doc-unix.html
  95. - Docs
  96. - More prominently, we should have a recommended apps list.
  97. - recommend gaim.
  98. - unrecommend IE because of ftp:// bug.
  99. N - we should add a preamble to tor-design saying it's out of date.
  100. N - Document transport and natdport
  101. - Improvements to bandwidth counting
  102. R - look into "uncounting" bytes spent on local connections, so
  103. we can bandwidthrate but still have fast downloads.
  104. R - "bandwidth classes", for incoming vs initiated-here conns,
  105. and to give dir conns lower priority.
  106. - Forward compatibility fixes
  107. - Caches should start trying to cache consensus docs?
  108. NR - Design
  109. N - Implement, if we think it's smart.
  110. - Start uploading short and long descriptors; authorities should support
  111. URLs to retrieve long descriptors, and should discard short descriptors
  112. for now. Later, once tools use the "long descriptor" URLs, authorities
  113. will serve the short descriptors every time they're asked for
  114. a descriptor.
  115. NR - Design
  116. N - Implement, if we think it's smart.
  117. - Check for any outstanding checks we do on the form or number of client
  118. certificates that would prevent us from executing certain
  119. blocking-resistance strategies.
  120. Topics to think about during 0.1.2.x development:
  121. * Figure out incentives.
  122. - (How can we make this tolerant of a bad v0?)
  123. * Figure out non-clique.
  124. * Figure out China.
  125. - Figure out partial network knowledge.
  126. - Figure out hidden services.
  127. - Design next-version protocol for directories
  128. - Design next-version protocol for connections
  129. Deferred from 0.1.2.x:
  130. - finish status event implementation and accompanying getinfos
  131. - More work on AvoidDiskWrites?
  132. NR- Get some kind of "meta signing key" to be used solely to sign
  133. releases/to certify releases when signed by the right people/
  134. to certify sign the right people's keys? Also use this to cert the SSL
  135. key, etc.
  136. - per-conn write buckets
  137. - separate config options for read vs write limiting
  138. (It's hard to support read > write, since we need better
  139. congestion control to avoid overfull buffers there. So,
  140. defer the whole thing.)
  141. P - Figure out why dll's compiled in mingw don't work right in WinXP.
  142. P - Figure out why openssl 0.9.8d "make test" fails at sha256t test.
  143. - don't do dns hijacking tests if we're reject *:* exit policy?
  144. (deferred until 0.1.1.x is less common)
  145. - Directory guards
  146. - RAM use in directory authorities.
  147. - Memory use improvements:
  148. - Look into pulling serverdescs off buffers as they arrive.
  149. - Save and mmap v1 directories, and networkstatus docs; store them
  150. zipped, not uncompressed.
  151. - Switch cached_router_t to use mmap.
  152. - What to do about reference counts on windows? (On Unix, this is
  153. easy: unlink works fine. (Right?) On Windows, I have doubts. Do we
  154. need to keep multiple files?)
  155. - What do we do about the fact that people can't read zlib-
  156. compressed files manually?
  157. - Change the way we handle cells, flow-control, and bridges.
  158. - The issue is that we package from edge connections aggressively until
  159. we hit their package windows or the circuit package windows, even if
  160. the buffer on the corresponding OR connection is pretty damn big. This
  161. sucks from a RAM usage POV. Now, we could try to stop reading on the
  162. edges (or just the edges connected to a local bridge) when an or_conn's
  163. outbuf is full. But if we're a server, and we stop reading on some
  164. exit conns when OR conns are full, soon OR conns will contain only
  165. traffic from other OR conns, and the exit data in question will never
  166. get written.
  167. - Remove socketpair-based bridges: use shared (or connected) buffers for
  168. communication, rather than sockets.
  169. - When relaying cells from an OR conn to an OR conn, have them wait in a
  170. queue on the or_circuit_t object; don't move them onto the target conn
  171. until there is space in the target conn's outbuf. Also, only package
  172. data from exitconns when there is space in the target conn's outbuf.
  173. - As an added advantage, this would let us kill stalled _circuits_
  174. when their buffers get too full, rather than killing entire OR
  175. conns. But we must think about anonymity implications of that.
  176. - We'll probably want to do some kind of big refactoring of our
  177. dataflow when we do these changes; stuff is hairy enough already,
  178. and it will only get harrier with this stuff.
  179. - If the client's clock is too far in the past, it will drop (or
  180. just not try to get) descriptors, so it'll never build circuits.
  181. - Tolerate clock skew on bridge relays.
  182. - A more efficient dir protocol.
  183. - Authorities should fetch the network-statuses amongst each
  184. other, consensus them, and advertise a communal network-status.
  185. This is not so much for safety/complexity as it is to reduce
  186. bandwidth requirements for Alice.
  187. - How does this interact with our goal of being able to choose
  188. your own dir authorities? I guess we're now assuming that all
  189. dir authorities know all the other authorities in their "group"?
  190. - Should we also look into a "delta since last network-status
  191. checkpoint" scheme, to reduce overhead further?
  192. - Extend the "r" line in network-status to give a set of buckets (say,
  193. comma-separated) for that router.
  194. - Buckets are deterministic based on IP address.
  195. - Then clients can choose a bucket (or set of buckets) to
  196. download and use.
  197. - Improvements to versioning.
  198. - When we connect to a Tor server, it sends back a cell listing
  199. the IP it believes it is using. Use this to block dvorak's attack.
  200. Also, this is a fine time to say what time you think it is.
  201. o Verify that a new cell type is okay with deployed codebase
  202. . Specify HELLO cells
  203. . Figure out v0 compatibility.
  204. - Implement
  205. - Eventdns improvements
  206. - Have a way to query for AAAA and A records simultaneously.
  207. - Improve request API: At the very least, add the ability to construct
  208. a more-or-less arbitrary request and get a response.
  209. - (Can we suppress cnames? Should we?)
  210. - Now that we're avoiding exits when picking non-exit positions,
  211. we need to consider how to pick nodes for internal circuits. If
  212. we avoid exits for all positions, we skew the load balancing. If
  213. we accept exits for all positions, we leak whether it's an internal
  214. circuit at every step. If we accept exits only at the last hop, we
  215. reintroduce Lasse's attacks from the Oakland paper.
  216. - We should ship with a list of stable dir mirrors -- they're not
  217. trusted like the authorities, but they'll provide more robustness
  218. and diversity for bootstrapping clients.
  219. - Simplify authority operation
  220. - Follow weasel's proposal, crossed with mixminion dir config format
  221. - A way to adjust router flags from the controller.
  222. (How do we prevent the authority from clobbering them soon after?)
  223. - a way to pick entry guards based wholly on extend_info equivalent;
  224. a way to export extend_info equivalent.
  225. - Better estimates in the directory of whether servers have good uptime
  226. (high expected time to failure) or good guard qualities (high
  227. fractional uptime).
  228. - AKA Track uptime as %-of-time-up, as well as time-since-last-down
  229. - Have a "Faster" status flag that means it. Fast2, Fast4, Fast8?
  230. - spec
  231. - implement
  232. - Failed rend desc fetches sometimes don't get retried. True/false?
  233. - Windows server usability
  234. - Solve the ENOBUFS problem.
  235. - make tor's use of openssl operate on buffers rather than sockets,
  236. so we can make use of libevent's buffer paradigm once it has one.
  237. - make tor's use of libevent tolerate either the socket or the
  238. buffer paradigm; includes unifying the functions in connect.c.
  239. - We need a getrlimit equivalent on Windows so we can reserve some
  240. file descriptors for saving files, etc. Otherwise we'll trigger
  241. asserts when we're out of file descriptors and crash.
  242. M - rewrite how libevent does select() on win32 so it's not so very slow.
  243. - Add overlapped IO
  244. - Add an option (related to AvoidDiskWrites) to disable directory caching.
  245. - More status event features:
  246. - Missing events:
  247. - DIR_REACHABLE
  248. - BAD_DIR_RESPONSE (Unexpected directory response; maybe we're behind
  249. a firewall.)
  250. - BAD_PROXY (Bad http or https proxy)
  251. - UNRECOGNIZED_ROUTER (a nickname we asked for is unavailable)
  252. - Status events related to hibernation
  253. - something about failing to parse our address?
  254. from resolve_my_address() in config.c
  255. - sketchy OS, sketchy threading
  256. - too many onions queued: threading problems or slow CPU?
  257. - Missing fields:
  258. - TIMEOUT on CHECKING_REACHABILITY
  259. - GETINFO status/client, status/server, status/general: There should be
  260. some way to learn which status events are currently "in effect."
  261. We should specify which these are, what format they appear in, and so
  262. on.
  263. Minor items for 0.1.2.x as time permits:
  264. - include bandwidth breakdown by conn->type in BW events.
  265. - getinfo ns/name/moria2 doesn't include a "v" line, even when some
  266. network-statuses I have show it. I suppose the fix should go in
  267. networkstatus_getinfo_helper_single() in routerlist.c.
  268. - Unify autoconf search code for libevent and openssl. Make code
  269. suggest platform-appropriate "devel" / "dev" / whatever packages
  270. if we can link but we can't find the headers.
  271. - Recommend polipo? Please?
  272. - Make documentation realize that location of system configuration file
  273. will depend on location of system defaults, and isn't always /etc/torrc.
  274. - Review torrc.sample to make it more discursive.
  275. o when reporting clock skew (both to logs and to controller), if it's
  276. taken 126 seconds to read from the directory, our clock skew estimate
  277. is 126 seconds wrong. use conn->timestamp_create or _lastwritten
  278. for a closer estimate?
  279. - a way to generate the website diagrams from source, so we can
  280. translate them as utf-8 text rather than with gimp.
  281. R - add d64 and fp64 along-side d and fp so people can paste status
  282. entries into a url. since + is a valid base64 char, only allow one
  283. at a time. spec and then do.
  284. - When we export something from foo.c file for testing purposes only,
  285. make a foo_test.h file for test.c to include.
  286. - The Debian package now uses --verify-config when (re)starting,
  287. to distinguish configuration errors from other errors. Perhaps
  288. the RPM and other startup scripts should too?
  289. - add a "default.action" file to the tor/vidalia bundle so we can fix the
  290. https thing in the default configuration:
  291. http://wiki.noreply.org/noreply/TheOnionRouter/TorFAQ#PrivoxyWeirdSSLPort
  292. . Flesh out options_description array in src/or/config.c
  293. - Don't let 'newnym' be triggered more often than every n seconds.
  294. X If we try to publish as a nickname that's already claimed, should
  295. we append a number (or increment the number) and try again? This
  296. way people who read their logs can fix it as before, but people
  297. who don't read their logs will still offer Tor servers.
  298. - Fall back to unnamed; warn user; send controller event. ("When we
  299. notice a 'Rejected: There is already a named server with this nickname'
  300. message... or maybe instead when we see in the networkstatuses that
  301. somebody else is Named with the name we want: warn the user, send a
  302. STATUS_SERVER message, and fall back to unnamed.")
  303. ! - Tor should bind its ports before dropping privs, so users don't
  304. have to do the ipchains dance.
  305. - Rate limit exit connections to a given destination -- this helps
  306. us play nice with websites when Tor users want to crawl them; it
  307. also introduces DoS opportunities.
  308. - Streamline how we pick entry nodes: Make choose_random_entry() have
  309. less magic and less control logic.
  310. - Christian Grothoff's attack of infinite-length circuit.
  311. the solution is to have a separate 'extend-data' cell type
  312. which is used for the first N data cells, and only
  313. extend-data cells can be extend requests.
  314. - Specify, including thought about anonymity implications.
  315. - Display the reasons in 'destroy' and 'truncated' cells under some
  316. circumstances?
  317. - We need a way for the authorities to declare that nodes are
  318. in a family. Also, it kinda sucks that family declarations use O(N^2)
  319. space in the descriptors.
  320. - If the server is spewing complaints about raising your ulimit -n,
  321. we should add a note about this to the server descriptor so other
  322. people can notice too.
  323. - cpu fixes:
  324. - see if we should make use of truncate to retry
  325. X kill dns workers more slowly
  326. . Directory changes
  327. . Some back-out mechanism for auto-approval
  328. - a way of rolling back approvals to before a timestamp
  329. - Consider minion-like fingerprint file/log combination.
  330. - packaging and ui stuff:
  331. . multiple sample torrc files
  332. . figure out how to make nt service stuff work?
  333. . Document it.
  334. - Vet all pending installer patches
  335. - Win32 installer plus privoxy, sockscap/freecap, etc.
  336. - Vet win32 systray helper code
  337. - Improve controller
  338. - a NEWSTATUS event similar to NEWDESC.
  339. - change circuit status events to give more details, like purpose,
  340. whether they're internal, when they become dirty, when they become
  341. too dirty for further circuits, etc.
  342. - What do we want here, exactly?
  343. - Specify and implement it.
  344. - Change stream status events analogously.
  345. - What do we want here, exactly?
  346. - Specify and implement it.
  347. - Make other events "better".
  348. - Change stream status events analogously.
  349. - What do we want here, exactly?
  350. - Specify and implement it.
  351. - Make other events "better" analogously
  352. - What do we want here, exactly?
  353. - Specify and implement it.
  354. . Expose more information via getinfo:
  355. - import and export rendezvous descriptors
  356. - Review all static fields for additional candidates
  357. - Allow EXTENDCIRCUIT to unknown server.
  358. - We need some way to adjust server status, and to tell tor not to
  359. download directories/network-status, and a way to force a download.
  360. - Make everything work with hidden services
  361. - Directory system improvements
  362. - config option to publish what ports you listen on, beyond
  363. ORPort/DirPort. It should support ranges and bit prefixes (?) too.
  364. - Parse this.
  365. - Relay this in networkstatus.
  366. - Be a DNS proxy.
  367. - Need a way to request address lookups (and allocate a stream ID for
  368. them) without having a corresponding client socket.
  369. - Once this is done, it would be nice to have a way to request address
  370. lookups from the controller without using SOCKS.
  371. - Refactor exit/entry side of DNS resolve: we don't need a connection_t;
  372. we can have an edge_connection_t and (say) dns_request_t both extend an
  373. edge_stream_t, and have p_streams and n_streams both be linked lists
  374. of edge_stream_t.
  375. - Look into generating torrc.{complete|sample}.in, tor.1.in,
  376. the HTML manual, and the online config documentation from a single
  377. source.
  378. - torrc.complete.in needs attention?
  379. Future version:
  380. - Configuration format really wants sections.
  381. - Good RBL substitute.
  382. - Authorities should try using exits for http to connect to some URLS
  383. (specified in a configuration file, so as not to make the List Of Things
  384. Not To Censor completely obvious) and ask them for results. Exits that
  385. don't give good answers should have the BadExit flag set.
  386. - Our current approach to block attempts to use Tor as a single-hop proxy
  387. is pretty lame; we should get a better one.
  388. . Update the hidden service stuff for the new dir approach.
  389. - switch to an ascii format, maybe sexpr?
  390. - authdirservers publish blobs of them.
  391. - other authdirservers fetch these blobs.
  392. - hidserv people have the option of not uploading their blobs.
  393. - you can insert a blob via the controller.
  394. - and there's some amount of backwards compatibility.
  395. - teach clients, intro points, and hidservs about auth mechanisms.
  396. - come up with a few more auth mechanisms.
  397. - auth mechanisms to let hidden service midpoint and responder filter
  398. connection requests.
  399. - Bind to random port when making outgoing connections to Tor servers,
  400. to reduce remote sniping attacks.
  401. - Have new people be in limbo and need to demonstrate usefulness
  402. before we approve them.
  403. - Clients should estimate their skew as median of skew from servers
  404. over last N seconds.
  405. - Make router_is_general_exit() a bit smarter once we're sure what it's for.
  406. - Audit everything to make sure rend and intro points are just as likely to
  407. be us as not.
  408. - Do something to prevent spurious EXTEND cells from making middleman
  409. nodes connect all over. Rate-limit failed connections, perhaps?
  410. - Automatically determine what ports are reachable and start using
  411. those, if circuits aren't working and it's a pattern we recognize
  412. ("port 443 worked once and port 9001 keeps not working").
  413. - Limit to 2 dir, 2 OR, N SOCKS connections per IP.
  414. - Handle full buffers without totally borking
  415. - Rate-limit OR and directory connections overall and per-IP and
  416. maybe per subnet.
  417. - Hold-open-until-flushed now works by accident; it should work by
  418. design.
  419. - DoS protection: TLS puzzles, public key ops, bandwidth exhaustion.
  420. - Specify?
  421. - tor-resolve script should use socks5 to get better error messages.
  422. - hidserv offerers shouldn't need to define a SocksPort
  423. * figure out what breaks for this, and do it.
  424. - tor should be able to have a pool of outgoing IP addresses
  425. that it is able to rotate through. (maybe)
  426. - Specify; implement.
  427. - let each hidden service (or other thing) specify its own
  428. OutboundBindAddress?
  429. - Stop using tor_socketpair to make connection bridges: do an
  430. implementation that uses buffers only.
  431. Blue-sky:
  432. - Patch privoxy and socks protocol to pass strings to the browser.
  433. - Standby/hotswap/redundant hidden services.
  434. - Robust decentralized storage for hidden service descriptors.
  435. - The "China problem"
  436. - Allow small cells and large cells on the same network?
  437. - Cell buffering and resending. This will allow us to handle broken
  438. circuits as long as the endpoints don't break, plus will allow
  439. connection (tls session key) rotation.
  440. - Implement Morphmix, so we can compare its behavior, complexity, etc.
  441. - Other transport. HTTP, udp, rdp, airhook, etc. May have to do our own
  442. link crypto, unless we can bully openssl into it.
  443. - Need a relay teardown cell, separate from one-way ends.
  444. (Pending a user who needs this)
  445. - Handle half-open connections: right now we don't support all TCP
  446. streams, at least according to the protocol. But we handle all that
  447. we've seen in the wild.
  448. (Pending a user who needs this)
  449. Non-Coding:
  450. - Mark up spec; note unclear points about servers
  451. - Mention controller libs someplace.
  452. . more pictures from ren. he wants to describe the tor handshake
  453. NR- write a spec appendix for 'being nice with tor'
  454. - tor-in-the-media page
  455. - Remove need for HACKING file.
  456. - Figure out licenses for website material.
  457. - Specify the keys and key rotation schedules and stuff
  458. Website:
  459. - and remove home and make the "Tor" picture be the link to home.
  460. - put the logo on the website, in source form, so people can put it on
  461. stickers directly, etc.
  462. R - make a page with the hidden service diagrams.
  463. - ask Jan to be the translation coordinator? add to volunteer page.
  464. - add a page for localizing all tor's components.
  465. - It would be neat if we had a single place that described _all_ the
  466. tor-related tools you can use, and what they give you, and how well they
  467. work. Right now, we don't give a lot of guidance wrt
  468. torbutton/foxproxy/privoxy/polipo in any consistent place.