TODO 16 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358
  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. Important bugfixes in 0.1.2.x:
  19. - 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. - 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. Items for 0.1.2.x:
  26. - when we start, remove any entryguards that are listed in excludenodes.
  27. . start calling dev releases 0.1.2.1-alpha-dev, not -cvs. Do we need
  28. to change the code in any way for this?
  29. - enumerate events of important things that occur in tor, so vidalia can
  30. react.
  31. - We should ship with a list of stable dir mirrors -- they're not
  32. trusted like the authorities, but they'll provide more robustness
  33. and diversity for bootstrapping clients.
  34. - Servers are easy to setup and run: being a relay is about as easy as
  35. being a client.
  36. - Reduce resource load
  37. - look into "uncounting" bytes spent on local connections. so
  38. we can bandwidthrate but still have fast downloads.
  39. - Write limiting; separate token bucket for write
  40. o dir answers include a your-ip-address-is header, so we can
  41. break our dependency on dyndns.
  42. - Come up with good 'nicknames' automatically, or make no-nickname
  43. routers workable.
  44. - tolerate clock skew on bridge relays.
  45. - a way to export server descriptors to controllers, and/or to
  46. non-standard dir authorities.
  47. - a way to pick entries based wholly on extend_info equivalent;
  48. a way to export extend_info equivalent.
  49. - option to dl directory info via tor
  50. - is the __AllDirActionsPrivate config option sufficient?
  51. D Count TLS bandwidth more accurately
  52. - Write-limit directory responses (need to research)
  53. N . Improve memory usage on tight-memory machines.
  54. . Directory-related fixes.
  55. . Mmap cache files where possible.
  56. o Mmap cached-routers file; when building it, go oldest-to-newest.
  57. - More unit tests and asserts for cached-routers file: ensure digest
  58. for the right router. Verify dl by digest, fp, etc.
  59. o Make sure cached-routers values and offsets are correct in the
  60. presence of windows FS insanity.
  61. D Save and mmap v1 directories, and networkstatus docs; store them
  62. zipped, not uncompressed.
  63. D Switch cached_router_t to use mmap.
  64. D What to do about reference counts on windows?
  65. (On Unix, this is easy: unlink works fine. (Right?) On
  66. Windows, I have doubts. Do we need to keep multiple files?)
  67. D What do we do about the fact that people can't read zlib-
  68. compressed files manually?
  69. - Look into pulling serverdescs off buffers as they arrive.
  70. - "bandwidth classes", for incoming vs initiated-here conns.
  71. o Asynchronous DNS
  72. o And test it
  73. . Make it work on windows.
  74. o Implement
  75. . Enable
  76. - Test
  77. - Make the Nameservers option documented, and make it work right on
  78. reload.
  79. o Fail when we have no configured nameservers!
  80. o Make it the default on platforms where it works.
  81. - Add ipv6 support; make API closer to getaddrinfo(). (i.e., allow a
  82. single AAAA/A query, return cname as well)
  83. - Teach it to be able to listen for A and PTR requests to be processed.
  84. Interface should be set_request_listener(sock, cb); [ cb(request) ]
  85. send_reply(request, answer);.
  86. - Refactor exit side of resolve: do we need a connection_t?
  87. - Refactor entry side of resolve: do we need a connection_t?
  88. - Security improvements
  89. - Directory guards
  90. - remember the last time we saw one of our entry guards labelled with
  91. the GUARD flag. If it's been too long, it is not suitable for use.
  92. If it's been really too long, remove it from the list.
  93. . Figure out avoiding duplicate /24 lines
  94. o automatically add /16 servers to family
  95. - do it in an efficient way. keep a list of something somewhere?
  96. - make the "16" part configurable, so people who run their own
  97. tor network can set it to 32.
  98. - Make reverse DNS work.
  99. - Specify
  100. X Implement with dnsworkers
  101. (There's no point doing this, since we will throw away dnsworkers once
  102. eventdns is confirmed to work everywhere.)
  103. o Implement in eventdns
  104. - Connect to resolve cells, server-side.
  105. - Add client-side interface
  106. - Performance improvements
  107. - Better estimates in the directory of whether servers have good uptime
  108. (high expected time to failure) or good guard qualities (high
  109. fractional uptime).
  110. - AKA Track uptime as %-of-time-up, as well as time-since-last-down.
  111. - Clients should prefer to avoid exit nodes for non-exit path positions.
  112. (bug 200)
  113. - Have a "Faster" status flag that means it. Fast2, Fast4, Fast8?
  114. - A more efficient dir protocol.
  115. N - Clients stop dumping old descriptors if the network-statuses
  116. claim they're still valid.
  117. - Later, servers will stop generating new descriptors simply
  118. because 18 hours have passed.
  119. - Authorities should fetch the network-statuses amongst each
  120. other, consensus them, and advertise a communal network-status.
  121. This is not so much for safety/complexity as it is to reduce
  122. bandwidth requirements for Alice.
  123. - How does this interact with our goal of being able to choose
  124. your own dir authorities? I guess we're now assuming that all
  125. dir authorities know all the other authorities in their "group"?
  126. - Should we also look into a "delta since last network-status
  127. checkpoint" scheme, to reduce overhead further?
  128. D Extend the "r" line in network-status to give a set of buckets (say,
  129. comma-separated) for that router.
  130. - Buckets are deterministic based on IP address.
  131. - Then clients can choose a bucket (or set of buckets) to
  132. download and use.
  133. - Critical but minor bugs, backport candidates.
  134. - Failed rend desc fetches sometimes don't get retried. True/false?
  135. . If we fail to connect via an exit enclave, (warn and) try again
  136. without demanding that exit node.
  137. - And recognize when extending to the enclave node is failing,
  138. so we can abandon then too.
  139. - non-v1 authorities should not accept rend descs.
  140. - We need a separate list of "hidserv authorities" if we want to
  141. retire moria1 from the main list.
  142. - support dir 503s better
  143. o clients don't log as loudly when they receive them
  144. - they don't count toward the 3-strikes rule
  145. - should there be some threshold of 503's after which we give up?
  146. - think about how to split "router is down" from "dirport shouldn't
  147. be tried for a while"?
  148. - authorities should *never* 503 a cache, but *should* 503 clients
  149. when they feel like it.
  150. - update dir-spec with what we decided for each of these
  151. - provide no-cache no-index headers from the dirport?
  152. - Windows server usability
  153. - Solve the ENOBUFS problem.
  154. - make tor's use of openssl operate on buffers rather than sockets,
  155. so we can make use of libevent's buffer paradigm once it has one.
  156. - make tor's use of libevent tolerate either the socket or the
  157. buffer paradigm; includes unifying the functions in connect.c.
  158. - We need a getrlimit equivalent on Windows so we can reserve some
  159. file descriptors for saving files, etc. Otherwise we'll trigger
  160. asserts when we're out of file descriptors and crash.
  161. M - rewrite how libevent does select() on win32 so it's not so very slow.
  162. - Add overlapped IO
  163. N - When we connect to a Tor server, it sends back a cell listing
  164. the IP it believes it is using. Use this to block dvorak's attack.
  165. Also, this is a fine time to say what time you think it is.
  166. o Verify that a new cell type is okay with deployed codebase
  167. . Specify HELLO cells
  168. . Figure out v0 compatibility.
  169. - Implement
  170. N - Exitlist should avoid outputting the same IP address twice.
  171. N - Write path-spec.txt
  172. - Packaging
  173. - Tell people about OSX Uninstaller
  174. - Quietly document NT Service options
  175. - Docs
  176. - More prominently, we should have a recommended apps list.
  177. - recommend gaim.
  178. - unrecommend IE because of ftp:// bug.
  179. - torrc.complete.in needs attention?
  180. Topics to think about during 0.1.2.x development:
  181. * Figure out incentives.
  182. - (How can we make this tolerant of a bad v0?)
  183. * Figure out non-clique.
  184. * Figure out China.
  185. - Figure out partial network knowledge.
  186. - Figure out hidden services.
  187. Minor items for 0.1.2.x as time permits.
  188. - Tor should bind its ports before dropping privs, so users don't
  189. have to do the ipchains dance.
  190. - Make --verify-config return a useful error code.
  191. - Rate limit exit connections to a given destination -- this helps
  192. us play nice with websites when Tor users want to crawl them; it
  193. also introduces DoS opportunities.
  194. - The bw_accounting file should get merged into the state file.
  195. - Streamline how we define a guard node as 'up'.
  196. - Better installers and build processes.
  197. - Commit edmanm's win32 makefile to tor contrib, or write a new one.
  198. - Christian Grothoff's attack of infinite-length circuit.
  199. the solution is to have a separate 'extend-data' cell type
  200. which is used for the first N data cells, and only
  201. extend-data cells can be extend requests.
  202. - Specify, including thought about anonymity implications.
  203. - Display the reasons in 'destroy' and 'truncated' cells under some
  204. circumstances?
  205. - We need a way for the authorities to declare that nodes are
  206. in a family. Also, it kinda sucks that family declarations use O(N^2)
  207. space in the descriptors.
  208. - If the server is spewing complaints about raising your ulimit -n,
  209. we should add a note about this to the server descriptor so other
  210. people can notice too.
  211. - rate limit the number of exit connections to a given destination, to
  212. help with DoS/crawling issues.
  213. - cpu fixes:
  214. - see if we should make use of truncate to retry
  215. - kill dns workers more slowly
  216. . Directory changes
  217. . Some back-out mechanism for auto-approval
  218. - a way of rolling back approvals to before a timestamp
  219. - Consider minion-like fingerprint file/log combination.
  220. - packaging and ui stuff:
  221. . multiple sample torrc files
  222. . figure out how to make nt service stuff work?
  223. . Document it.
  224. - Vet all pending installer patches
  225. - Win32 installer plus privoxy, sockscap/freecap, etc.
  226. - Vet win32 systray helper code
  227. - Improve controller
  228. - change circuit status events to give more details, like purpose,
  229. whether they're internal, when they become dirty, when they become
  230. too dirty for further circuits, etc.
  231. - What do we want here, exactly?
  232. - Specify and implement it.
  233. - Change stream status events analogously.
  234. - What do we want here, exactly?
  235. - Specify and implement it.
  236. - Make other events "better".
  237. - Change stream status events analogously.
  238. - What do we want here, exactly?
  239. - Specify and implement it.
  240. - Make other events "better" analogously
  241. - What do we want here, exactly?
  242. - Specify and implement it.
  243. . Expose more information via getinfo:
  244. - import and export rendezvous descriptors
  245. - Review all static fields for additional candidates
  246. - Allow EXTENDCIRCUIT to unknown server.
  247. - We need some way to adjust server status, and to tell tor not to
  248. download directories/network-status, and a way to force a download.
  249. - It would be nice to request address lookups from the controller
  250. without using SOCKS.
  251. - Make everything work with hidden services
  252. - Directory system improvements
  253. - config option to publish what ports you listen on, beyond
  254. ORPort/DirPort. It should support ranges and bit prefixes (?) too.
  255. - Parse this.
  256. - Relay this in networkstatus.
  257. Future version:
  258. - Tor should have a "DNS port" so we don't need to ship with (and
  259. write) a clean portable dns proxy.
  260. . Update the hidden service stuff for the new dir approach.
  261. - switch to an ascii format, maybe sexpr?
  262. - authdirservers publish blobs of them.
  263. - other authdirservers fetch these blobs.
  264. - hidserv people have the option of not uploading their blobs.
  265. - you can insert a blob via the controller.
  266. - and there's some amount of backwards compatibility.
  267. - teach clients, intro points, and hidservs about auth mechanisms.
  268. - come up with a few more auth mechanisms.
  269. - auth mechanisms to let hidden service midpoint and responder filter
  270. connection requests.
  271. - Bind to random port when making outgoing connections to Tor servers,
  272. to reduce remote sniping attacks.
  273. - Have new people be in limbo and need to demonstrate usefulness
  274. before we approve them.
  275. - Clients should estimate their skew as median of skew from servers
  276. over last N seconds.
  277. - Make router_is_general_exit() a bit smarter once we're sure what it's for.
  278. - Audit everything to make sure rend and intro points are just as likely to
  279. be us as not.
  280. - Do something to prevent spurious EXTEND cells from making middleman
  281. nodes connect all over. Rate-limit failed connections, perhaps?
  282. - Automatically determine what ports are reachable and start using
  283. those, if circuits aren't working and it's a pattern we recognize
  284. ("port 443 worked once and port 9001 keeps not working").
  285. - Limit to 2 dir, 2 OR, N SOCKS connections per IP.
  286. - Handle full buffers without totally borking
  287. - Rate-limit OR and directory connections overall and per-IP and
  288. maybe per subnet.
  289. - Hold-open-until-flushed now works by accident; it should work by
  290. design.
  291. - DoS protection: TLS puzzles, public key ops, bandwidth exhaustion.
  292. - Specify?
  293. - tor-resolve script should use socks5 to get better error messages.
  294. - hidserv offerers shouldn't need to define a SocksPort
  295. * figure out what breaks for this, and do it.
  296. - tor should be able to have a pool of outgoing IP addresses
  297. that it is able to rotate through. (maybe)
  298. - let each hidden service (or other thing) specify its own
  299. OutboundBindAddress?
  300. - Have a mode that doesn't write to disk much, so we can run Tor on
  301. flash memory (e.g. Linksys routers).
  302. Blue-sky:
  303. - Patch privoxy and socks protocol to pass strings to the browser.
  304. - Standby/hotswap/redundant hidden services.
  305. - Robust decentralized storage for hidden service descriptors.
  306. - The "China problem"
  307. - Allow small cells and large cells on the same network?
  308. - Cell buffering and resending. This will allow us to handle broken
  309. circuits as long as the endpoints don't break, plus will allow
  310. connection (tls session key) rotation.
  311. - Implement Morphmix, so we can compare its behavior, complexity, etc.
  312. - Other transport. HTTP, udp, rdp, airhook, etc. May have to do our own
  313. link crypto, unless we can bully openssl into it.
  314. - Need a relay teardown cell, separate from one-way ends.
  315. (Pending a user who needs this)
  316. - Handle half-open connections: right now we don't support all TCP
  317. streams, at least according to the protocol. But we handle all that
  318. we've seen in the wild.
  319. (Pending a user who needs this)
  320. Non-Coding:
  321. - Mark up spec; note unclear points about servers
  322. - Mention controller libs someplace.
  323. P - flesh out the rest of the section 6 of the faq
  324. . more pictures from ren. he wants to describe the tor handshake
  325. NR- write a spec appendix for 'being nice with tor'
  326. - tor-in-the-media page
  327. - Remove need for HACKING file.
  328. - Figure out licenses for website material.
  329. Website:
  330. - and remove home and make the "Tor" picture be the link to home.
  331. - put the logo on the website, in source form, so people can put it on
  332. stickers directly, etc.
  333. R - make a page with the hidden service diagrams.
  334. - ask Jan to be the translation coordinator? add to volunteer page.