TODO 20 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451
  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 Deferred
  13. X Abandoned
  14. . <nickm> "Let's try to find a way to make it run and make the version
  15. match, but if not, let's just make it run."
  16. - <arma> "should we detect if we have a --with-ssl-dir and try the -R
  17. by default, if it works?"
  18. Items for 0.1.2.x, real soon now:
  19. x - When we've been idle a long time, we stop fetching server
  20. descriptors. When we then get a socks request, we build circuits
  21. immediately using whatever descriptors we have, rather than waiting
  22. until we've fetched correct ones.
  23. x - If the client's clock is too far in the past, it will drop (or
  24. just not try to get) descriptors, so it'll never build circuits.
  25. N - Bug 326: make eventdns thrash less.
  26. N - Test guard unreachable logic; make sure that we actually attempt to
  27. connect to guards that we think are unreachable from time to time.
  28. Make sure that we don't freak out when the network is down.
  29. N - Stop recommending exits as guards?
  30. N - Clients stop dumping old descriptors if the network-statuses
  31. claim they're still valid.
  32. P - Figure out why dll's compiled in mingw don't work right in Winxp.
  33. P - Figure out why openssl 0.9.8c "make test" fails at sha256t test.
  34. Items for 0.1.2.x:
  35. - Now that we're avoiding exits when picking non-exit positions,
  36. we need to consider how to pick nodes for internal circuits. If
  37. we avoid exits for all positions, we skew the load balancing. If
  38. we accept exits for all positions, we leak whether it's an internal
  39. circuit at every step. If we accept exits only at the last hop, we
  40. reintroduce Lasse's attacks from the Oakland paper.
  41. - enumerate events of important things that occur in tor, so vidalia can
  42. react.
  43. N - Backend implementation
  44. R - Actually list all the events (notice and warn log messages are a good
  45. place to look.) Divide messages into categories, perhaps.
  46. N - Specify general event system
  47. R - Specify actual events.
  48. N . Have (and document) a BEGIN_DIR relay cell that means "Connect to your
  49. directory port."
  50. o Specify
  51. o Implement
  52. - Use for something, so we can be sure it works.
  53. - Test and debug
  54. N - Send back RELAY_END cells on malformed RELAY_BEGIN.
  55. N - Change the circuit end reason display a little for reasons from
  56. destroyed/truncated circuits. We want to indicate both that we're
  57. closing because somebody told us to, and why they told us they wanted to
  58. close.
  59. x - We should ship with a list of stable dir mirrors -- they're not
  60. trusted like the authorities, but they'll provide more robustness
  61. and diversity for bootstrapping clients.
  62. N - Simplify authority operation
  63. - Follow weasel's proposal, crossed with mixminion dir config format
  64. - Reject/invalidate by IP.
  65. - Servers are easy to setup and run: being a relay is about as easy as
  66. being a client.
  67. . Reduce resource load
  68. d - Tolerate clock skew on bridge relays.
  69. N - A way to examine router flags from controller.
  70. d - A way to adjust router flags from the controller
  71. d - a way to pick entries based wholly on extend_info equivalent;
  72. a way to export extend_info equivalent.
  73. R - option to dl directory info via tor
  74. - Make an option like __AllDirActionsPrivate that falls back to
  75. non-Tor DL when not enough info present.
  76. D Count TLS bandwidth more accurately
  77. - Improvements to bandwidth counting
  78. R - look into "uncounting" bytes spent on local connections, so
  79. we can bandwidthrate but still have fast downloads.
  80. R - "bandwidth classes", for incoming vs initiated-here conns.
  81. d - Write limiting; separate token bucket for write
  82. - Write-limit directory responses (need to research)
  83. N - DNS improvements
  84. o Option to deal with broken DNS of the "ggoogle.com? Ah, you meant
  85. ads.me.com!" variety.
  86. o Autodetect whether DNS is broken in this way.
  87. - Additional fix: allow clients to have some addresses that mean,
  88. notfound. Yes, this blacklists IPs for having ever been used by
  89. DNS hijackers.
  90. o Don't ask reject *:* nodes for DNS unless client wants you to.
  91. . Asynchronous DNS
  92. o Document and rename SearchDomains, ResolvConf options
  93. D Make API closer to getaddrinfo()
  94. - Teach it to be able to listen for A and PTR requests to be processed.
  95. Interface should be set_request_listener(sock, cb); [ cb(request) ]
  96. send_reply(request, answer);
  97. d - Add option to use /etc/hosts?
  98. d - Special-case localhost?
  99. - Verify that it works on windows
  100. . Make reverse DNS work.
  101. o Specify
  102. X Implement with dnsworkers
  103. (There's no point doing this, since we will throw away dnsworkers once
  104. eventdns is confirmed to work everywhere.)
  105. o Implement in eventdns
  106. o Connect to resolve cells, server-side.
  107. o Add element to routerinfo to note routers that aren't using eventdns,
  108. so we can avoid sending them reverse DNS etc.
  109. o Fix the bug with server-side caching, whatever is causing it.
  110. . Add client-side interface
  111. o SOCKS interface: specify
  112. o SOCKS interface: implement
  113. - Cache answers client-side
  114. o Add to Tor-resolve.py
  115. - Add to tor-resolve
  116. - Check for invalid characters in hostnames before trying to resolve
  117. them. (This will help catch attempts do to mean things to our DNS
  118. server, and bad software that tries to do DNS lookups on whole URLs.)
  119. - address_is_invalid_destination() is the right thing to call here
  120. (and feel free to make that function smarter)
  121. - Performance improvements
  122. x - Better estimates in the directory of whether servers have good uptime
  123. (high expected time to failure) or good guard qualities (high
  124. fractional uptime).
  125. - AKA Track uptime as %-of-time-up, as well as time-since-last-down
  126. - Have a "Faster" status flag that means it. Fast2, Fast4, Fast8?
  127. x - spec
  128. d - implement
  129. - A more efficient dir protocol.
  130. N - Later, servers will stop generating new descriptors simply
  131. because 18 hours have passed: we must start tolerating this now.
  132. - Critical but minor bugs, backport candidates.
  133. d - Failed rend desc fetches sometimes don't get retried. True/false?
  134. R - support dir 503s better
  135. o clients don't log as loudly when they receive them
  136. - they don't count toward the 3-strikes rule
  137. - should there be some threshold of 503's after which we give up?
  138. - Delay when we get a lot of 503s.
  139. N - split "router is down" from "dirport shouldn't be tried for a while"?
  140. Just a separate bit.
  141. - authorities should *never* 503 a cache, but *should* 503 clients
  142. when they feel like it.
  143. - update dir-spec with what we decided for each of these
  144. N - provide no-cache no-index headers from the dirport?
  145. - Specify
  146. - cacheing
  147. - Single network-statuses, single descriptors, "all", "authority",
  148. and v1 directory stuff are all cacheable for a short time.
  149. - Multiple network-statuses or descriptors are not cacheable.
  150. - Be sure to be correct wrt HTTP/1.0
  151. - indexing
  152. - robots.txt
  153. - Implement
  154. - Windows server usability
  155. - Solve the ENOBUFS problem.
  156. - make tor's use of openssl operate on buffers rather than sockets,
  157. so we can make use of libevent's buffer paradigm once it has one.
  158. - make tor's use of libevent tolerate either the socket or the
  159. buffer paradigm; includes unifying the functions in connect.c.
  160. - We need a getrlimit equivalent on Windows so we can reserve some
  161. file descriptors for saving files, etc. Otherwise we'll trigger
  162. asserts when we're out of file descriptors and crash.
  163. M - rewrite how libevent does select() on win32 so it's not so very slow.
  164. - Add overlapped IO
  165. Nd- Have a mode that doesn't write to disk much, so we can run Tor on
  166. flash memory (e.g. Linksys routers or USB keys).
  167. o Add AvoidDiskWrites config option.
  168. - only write state file when it's "changed"
  169. - stop writing identity key / fingerprint / etc every restart
  170. - stop caching directory stuff -- and disable mmap?
  171. - more?
  172. NR. Write path-spec.txt
  173. - Packaging
  174. - Tell people about OSX Uninstaller
  175. - Quietly document NT Service options
  176. - Switch canonical win32 compiler to mingw.
  177. NR - Get some kind of "meta signing key" to be used solely to sign
  178. releases/to certify releases when signed by the right people/
  179. to certify sign the right people's keys? Also use this to cert the SSL
  180. key, etc.
  181. - If we haven't replaced privoxy, lock down its configuration in all
  182. packages, as documented in tor-doc-unix.html
  183. - Docs
  184. - More prominently, we should have a recommended apps list.
  185. - recommend gaim.
  186. - unrecommend IE because of ftp:// bug.
  187. - torrc.complete.in needs attention?
  188. - we should add a preamble to tor-design saying it's out of date.
  189. Topics to think about during 0.1.2.x development:
  190. * Figure out incentives.
  191. - (How can we make this tolerant of a bad v0?)
  192. * Figure out non-clique.
  193. * Figure out China.
  194. - Figure out partial network knowledge.
  195. - Figure out hidden services.
  196. - Design next-version protocol for directories
  197. - Design next-version protocol for connections
  198. For blocking-resistance scheme:
  199. X allow ordinary-looking ssl for dir connections. need a new dirport
  200. for this, or can we handle both ssl and non-ssl, or should we
  201. entirely switch to ssl in certain cases?
  202. d - need to figure out how to fetch status of a few servers from the BDA
  203. without fetching all statuses. A new URL to fetch I presume?
  204. Deferred from 0.1.2.x:
  205. - Directory guards
  206. - RAM use in directory authorities.
  207. - Memory use improvements:
  208. - Look into pulling serverdescs off buffers as they arrive.
  209. - Save and mmap v1 directories, and networkstatus docs; store them
  210. zipped, not uncompressed.
  211. - Switch cached_router_t to use mmap.
  212. - What to do about reference counts on windows? (On Unix, this is
  213. easy: unlink works fine. (Right?) On Windows, I have doubts. Do we
  214. need to keep multiple files?)
  215. - What do we do about the fact that people can't read zlib-
  216. compressed files manually?
  217. - Add IPv6 support to eventdns.c
  218. - Refactor DNS resolve implementation
  219. - Refactor exit side of resolve: do we need a connection_t?
  220. - Refactor entry side of resolve: do we need a connection_t?
  221. - A more efficient dir protocol.
  222. - Authorities should fetch the network-statuses amongst each
  223. other, consensus them, and advertise a communal network-status.
  224. This is not so much for safety/complexity as it is to reduce
  225. bandwidth requirements for Alice.
  226. - How does this interact with our goal of being able to choose
  227. your own dir authorities? I guess we're now assuming that all
  228. dir authorities know all the other authorities in their "group"?
  229. - Should we also look into a "delta since last network-status
  230. checkpoint" scheme, to reduce overhead further?
  231. - Extend the "r" line in network-status to give a set of buckets (say,
  232. comma-separated) for that router.
  233. - Buckets are deterministic based on IP address.
  234. - Then clients can choose a bucket (or set of buckets) to
  235. download and use.
  236. - Improvements to versioning.
  237. - When we connect to a Tor server, it sends back a cell listing
  238. the IP it believes it is using. Use this to block dvorak's attack.
  239. Also, this is a fine time to say what time you think it is.
  240. o Verify that a new cell type is okay with deployed codebase
  241. . Specify HELLO cells
  242. . Figure out v0 compatibility.
  243. - Implement
  244. Minor items for 0.1.2.x as time permits:
  245. - The Debian package now uses --verify-config when (re)starting,
  246. to distinguish configuration errors from other errors. Perhaps
  247. the RPM and other startup scripts should too?
  248. - add a "default.action" file to the tor/vidalia bundle so we can fix the
  249. https thing in the default configuration:
  250. http://wiki.noreply.org/noreply/TheOnionRouter/TorFAQ#PrivoxyWeirdSSLPort
  251. - even if your torrc lists yourself in your myfamily line, don't list it in
  252. the descriptor.
  253. - Flesh out options_description array in src/or/config.c
  254. - Don't let 'newnym' be triggered more often than every n seconds.
  255. - change log_fn() to log() on notice/warn/err logs where we can.
  256. - the deb now uses --verify-config to distinguish between configuration
  257. errors and other errors. Should the rpm, the ports, etc do this too?
  258. X If we try to publish as a nickname that's already claimed, should
  259. we append a number (or increment the number) and try again? This
  260. way people who read their logs can fix it as before, but people
  261. who don't read their logs will still offer Tor servers.
  262. - Fall back to unnamed; warn user; sent controller event.
  263. ! - Tor should bind its ports before dropping privs, so users don't
  264. have to do the ipchains dance.
  265. - Rate limit exit connections to a given destination -- this helps
  266. us play nice with websites when Tor users want to crawl them; it
  267. also introduces DoS opportunities.
  268. ! - The bw_accounting file should get merged into the state file.
  269. - Streamline how we pick entry nodes.
  270. ! - Better installers and build processes.
  271. - Commit edmanm's win32 makefile to tor contrib, or write a new one.
  272. - Christian Grothoff's attack of infinite-length circuit.
  273. the solution is to have a separate 'extend-data' cell type
  274. which is used for the first N data cells, and only
  275. extend-data cells can be extend requests.
  276. - Specify, including thought about anonymity implications.
  277. - Display the reasons in 'destroy' and 'truncated' cells under some
  278. circumstances?
  279. - We need a way for the authorities to declare that nodes are
  280. in a family. Also, it kinda sucks that family declarations use O(N^2)
  281. space in the descriptors.
  282. - If the server is spewing complaints about raising your ulimit -n,
  283. we should add a note about this to the server descriptor so other
  284. people can notice too.
  285. - cpu fixes:
  286. - see if we should make use of truncate to retry
  287. X kill dns workers more slowly
  288. . Directory changes
  289. . Some back-out mechanism for auto-approval
  290. - a way of rolling back approvals to before a timestamp
  291. - Consider minion-like fingerprint file/log combination.
  292. - packaging and ui stuff:
  293. . multiple sample torrc files
  294. . figure out how to make nt service stuff work?
  295. . Document it.
  296. - Vet all pending installer patches
  297. - Win32 installer plus privoxy, sockscap/freecap, etc.
  298. - Vet win32 systray helper code
  299. - Improve controller
  300. - a NEWSTATUS event similar to NEWDESC.
  301. - change circuit status events to give more details, like purpose,
  302. whether they're internal, when they become dirty, when they become
  303. too dirty for further circuits, etc.
  304. - What do we want here, exactly?
  305. - Specify and implement it.
  306. - Change stream status events analogously.
  307. - What do we want here, exactly?
  308. - Specify and implement it.
  309. - Make other events "better".
  310. - Change stream status events analogously.
  311. - What do we want here, exactly?
  312. - Specify and implement it.
  313. - Make other events "better" analogously
  314. - What do we want here, exactly?
  315. - Specify and implement it.
  316. . Expose more information via getinfo:
  317. - import and export rendezvous descriptors
  318. - Review all static fields for additional candidates
  319. - Allow EXTENDCIRCUIT to unknown server.
  320. - We need some way to adjust server status, and to tell tor not to
  321. download directories/network-status, and a way to force a download.
  322. - It would be nice to request address lookups from the controller
  323. without using SOCKS.
  324. - Make everything work with hidden services
  325. - Directory system improvements
  326. - config option to publish what ports you listen on, beyond
  327. ORPort/DirPort. It should support ranges and bit prefixes (?) too.
  328. - Parse this.
  329. - Relay this in networkstatus.
  330. Future version:
  331. - Configuration format really wants sections.
  332. - Good RBL substitute.
  333. - Authorities should try using exits for http to connect to some URLS
  334. (specified in a configuration file, so as not to make the List Of Things
  335. Not To Censor completely obvious) and ask them for results. Exits that
  336. don't give good answers should have the BadExit flag set.
  337. - Our current approach to block attempts to use Tor as a single-hop proxy
  338. is pretty lame; we should get a better one.
  339. . Update the hidden service stuff for the new dir approach.
  340. - switch to an ascii format, maybe sexpr?
  341. - authdirservers publish blobs of them.
  342. - other authdirservers fetch these blobs.
  343. - hidserv people have the option of not uploading their blobs.
  344. - you can insert a blob via the controller.
  345. - and there's some amount of backwards compatibility.
  346. - teach clients, intro points, and hidservs about auth mechanisms.
  347. - come up with a few more auth mechanisms.
  348. - auth mechanisms to let hidden service midpoint and responder filter
  349. connection requests.
  350. - Bind to random port when making outgoing connections to Tor servers,
  351. to reduce remote sniping attacks.
  352. - Have new people be in limbo and need to demonstrate usefulness
  353. before we approve them.
  354. - Clients should estimate their skew as median of skew from servers
  355. over last N seconds.
  356. - Make router_is_general_exit() a bit smarter once we're sure what it's for.
  357. - Audit everything to make sure rend and intro points are just as likely to
  358. be us as not.
  359. - Do something to prevent spurious EXTEND cells from making middleman
  360. nodes connect all over. Rate-limit failed connections, perhaps?
  361. - Automatically determine what ports are reachable and start using
  362. those, if circuits aren't working and it's a pattern we recognize
  363. ("port 443 worked once and port 9001 keeps not working").
  364. - Limit to 2 dir, 2 OR, N SOCKS connections per IP.
  365. - Handle full buffers without totally borking
  366. - Rate-limit OR and directory connections overall and per-IP and
  367. maybe per subnet.
  368. - Hold-open-until-flushed now works by accident; it should work by
  369. design.
  370. - DoS protection: TLS puzzles, public key ops, bandwidth exhaustion.
  371. - Specify?
  372. - tor-resolve script should use socks5 to get better error messages.
  373. - hidserv offerers shouldn't need to define a SocksPort
  374. * figure out what breaks for this, and do it.
  375. - tor should be able to have a pool of outgoing IP addresses
  376. that it is able to rotate through. (maybe)
  377. - Specify; implement.
  378. - let each hidden service (or other thing) specify its own
  379. OutboundBindAddress?
  380. - Stop using tor_socketpair to make connection bridges: do an
  381. implementation that uses buffers only.
  382. Blue-sky:
  383. - Patch privoxy and socks protocol to pass strings to the browser.
  384. - Standby/hotswap/redundant hidden services.
  385. - Robust decentralized storage for hidden service descriptors.
  386. - The "China problem"
  387. - Allow small cells and large cells on the same network?
  388. - Cell buffering and resending. This will allow us to handle broken
  389. circuits as long as the endpoints don't break, plus will allow
  390. connection (tls session key) rotation.
  391. - Implement Morphmix, so we can compare its behavior, complexity, etc.
  392. - Other transport. HTTP, udp, rdp, airhook, etc. May have to do our own
  393. link crypto, unless we can bully openssl into it.
  394. - Need a relay teardown cell, separate from one-way ends.
  395. (Pending a user who needs this)
  396. - Handle half-open connections: right now we don't support all TCP
  397. streams, at least according to the protocol. But we handle all that
  398. we've seen in the wild.
  399. (Pending a user who needs this)
  400. Non-Coding:
  401. - Mark up spec; note unclear points about servers
  402. - Mention controller libs someplace.
  403. . more pictures from ren. he wants to describe the tor handshake
  404. NR- write a spec appendix for 'being nice with tor'
  405. - tor-in-the-media page
  406. - Remove need for HACKING file.
  407. - Figure out licenses for website material.
  408. Website:
  409. - and remove home and make the "Tor" picture be the link to home.
  410. - put the logo on the website, in source form, so people can put it on
  411. stickers directly, etc.
  412. R - make a page with the hidden service diagrams.
  413. - ask Jan to be the translation coordinator? add to volunteer page.