TODO 13 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301
  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. Non-Coding, Soon:
  15. N - Mark up spec; note unclear points about servers
  16. N - Mention controller libs someplace.
  17. D FAQ entry: why gnutls is bad/not good for tor
  18. P - flesh out the rest of the section 6 of the faq
  19. P - gather pointers to livecd distros that include tor
  20. R . more pictures from ren. he wants to describe the tor handshake, i want to
  21. talk about hidden services.
  22. NR- write a spec appendix for 'being nice with tor'
  23. - tor-in-the-media page
  24. - Remove need for HACKING file.
  25. Website:
  26. - and remove home and make the "Tor" picture be the link to home.
  27. - put the logo on the website, in source form, so people can put it on
  28. stickers directly, etc.
  29. for 0.1.1.x-rc:
  30. N - building on freebsd 6.0: (with multiple openssl installations)
  31. . <nickm> "Let's try to find a way to make it run and make the version
  32. match, but if not, let's just make it run."
  33. - <arma> should we detect if we have a --with-ssl-dir and try the -R
  34. by default, if it works?
  35. - support dir 503s better
  36. o clients don't log as loudly when they receive them
  37. - they don't count toward the 3-strikes rule
  38. - should there be some threshold of 503's after which we give up?
  39. - think about how to split "router is down" from "dirport shouldn't
  40. be tried for a while"?
  41. - authorities should *never* 503 a cache, but *should* 503 clients
  42. when they feel like it.
  43. - update dir-spec with what we decided for each of these
  44. o when logging unknown http headers, this could include bad escape codes?
  45. more generally, attacker-controller log entries with newlines in them
  46. are dangerous for our users.
  47. o So... add functions to escape potentially malicious values before
  48. logging them, and test values more closely as they arrive...
  49. - But what to do about contact_info and platform?
  50. - (Didn't finish converting rend*.c)
  51. - Make "setconf" and "hup" behavior cleaner for LINELIST config
  52. options (e.g. Log). Bug 238.
  53. R - Jan 26 10:25:04.832 [warn] add_an_entry_guard(): Tried finding a
  54. new entry, but failed. Bad news. XXX.
  55. R - streamline how we define a guard node as 'up'. document it
  56. somewhere.
  57. R - reduce log severity for guard nodes.
  58. R - make guard node timeout higher.
  59. o Clean and future-proof exit policy formats a bit.
  60. o Likewise accept, but don't generate /bits formats (unless they're
  61. accepted in 0.0.9 and later).
  62. o Warn when we see a netmask that isn't a prefix.
  63. o Make clients understand "private:*" in exit policies, even though
  64. we don't generate it yet.
  65. for 0.1.1.x-final:
  66. - find 10 dirservers.
  67. - Make it no longer default for v2 dirservers to support v1.
  68. - non-versioning dirservers don't need to set recommended*versions.
  69. - non-naming dirservers don't need to have an approved-routers file.
  70. - What are criteria to be a dirserver? Write a policy.
  71. - Document AuthDirInvalid, AuthDirReject, AuthDirRejectUnlisted
  72. - are there other options that we haven't documented so far?
  73. - look at the proposed os x uninstaller:
  74. http://archives.seul.org/or/talk/Jan-2006/msg00038.html
  75. Deferred from 0.1.1.x:
  76. N - commit edmanm's win32 makefile to tor cvs contrib
  77. R - look into "uncounting" bytes spent on local connections. so
  78. we can bandwidthrate but still have fast downloads.
  79. R - Christian Grothoff's attack of infinite-length circuit.
  80. the solution is to have a separate 'extend-data' cell type
  81. which is used for the first N data cells, and only
  82. extend-data cells can be extend requests.
  83. - Specify, including thought about
  84. - Implement
  85. R - When we connect to a Tor server, it sends back a signed cell listing
  86. the IP it believes it is using. Use this to block dvorak's attack.
  87. Also, this is a fine time to say what time you think it is.
  88. - Verify that a new cell type is okay with deployed codebase
  89. - Specify
  90. - Implement
  91. R - failed rend desc fetches sometimes don't get retried.
  92. - Add config options to not publish and not fetch rend descs.
  93. - Add controller interfaces to hear rend desc events and learn
  94. about rend descs. In base16 I guess for now.
  95. N - Display the reasons in 'destroy' and 'truncated' cells under some
  96. circumstances?
  97. - If the server is spewing complaints about raising your ulimit -n,
  98. we should add a note about this to the server descriptor so other
  99. people can notice too.
  100. - We need a getrlimit equivalent on Windows so we can reserve some
  101. file descriptors for saving files, etc. Otherwise we'll trigger
  102. asserts when we're out of file descriptors and crash.
  103. X <weasel> it would be nice to support a unix socket for the control thing.
  104. The main motivation behind this was that we could let unix permissions
  105. take care of the authentication step: everybody who can connect to the
  106. socket is authenticated. However, the linux unix(7) manual page suggests
  107. that requiring read/write permissions on the socket in order to use it
  108. is Linux specific, and that many BSD-derived systems ignore the permissions
  109. on the socket file. Portable programs should not rely on this feature for
  110. security, therefore the motivation for this feature is gone.
  111. - the tor client can do the "automatic proxy config url" thing?
  112. R - clients prefer to avoid exit nodes for non-exit path positions.
  113. - Automatically determine what ports are reachable and start using
  114. those, if circuits aren't working and it's a pattern we recognize
  115. ("port 443 worked once and port 9001 keeps not working").
  116. N - Should router info have a pointer to routerstatus?
  117. - We should at least do something about the duplicated fields.
  118. N . Additional controller features
  119. - change circuit status events to give more details, like purpose,
  120. whether they're internal, when they become dirty, when they become
  121. too dirty for further circuits, etc.
  122. R - What do we want here, exactly?
  123. N - Specify and implement it.
  124. - Change stream status events analogously.
  125. R - What do we want here, exactly?
  126. N - Specify and implement it.
  127. - Make other events "better".
  128. - Change stream status events analogously.
  129. R - What do we want here, exactly?
  130. N - Specify and implement it.
  131. - Make other events "better" analogously
  132. R - What do we want here, exactly?
  133. N - Specify and implement it.
  134. . Expose more information via getinfo:
  135. - import and export rendezvous descriptors
  136. - Review all static fields for additional candidates
  137. - Allow EXTENDCIRCUIT to unknown server.
  138. - We need some way to adjust server status, and to tell tor not to
  139. download directories/network-status, and a way to force a download.
  140. - It would be nice to request address lookups from the controller
  141. without using SOCKS.
  142. - Make everything work with hidden services
  143. X switch accountingmax to count total in+out, not either in or
  144. out. it's easy to move in this direction (not risky), but hard to
  145. back out if we decide we prefer it the way it already is. hm.
  146. - cpu fixes:
  147. - see if we should make use of truncate to retry
  148. R - kill dns workers more slowly
  149. . Directory changes
  150. . Some back-out mechanism for auto-approval
  151. - a way of rolling back approvals to before a timestamp
  152. - Consider minion-like fingerprint file/log combination.
  153. - config option to publish what ports you listen on, beyond
  154. ORPort/DirPort. It should support ranges and bit prefixes (?) too.
  155. - Parse this.
  156. - Relay this in networkstatus.
  157. - Non-directories don't need to keep descriptors in memory.
  158. o Make descriptor-fetching happen via an indirection function.
  159. - Remember file and offset.
  160. - Keep a journal FD for appending router descriptors.
  161. - packaging and ui stuff:
  162. . multiple sample torrc files
  163. - uninstallers
  164. . for os x
  165. . figure out how to make nt service stuff work?
  166. . Document it.
  167. . Add version number to directory.
  168. N - Vet all pending installer patches
  169. - Win32 installer plus privoxy, sockscap/freecap, etc.
  170. - Vet win32 systray helper code
  171. - document:
  172. - torcp needs more attention in the tor-doc-win32.
  173. - recommend gaim.
  174. - unrecommend IE because of ftp:// bug.
  175. - torrc.complete.in needs attention?
  176. - Bind to random port when making outgoing connections to Tor servers,
  177. to reduce remote sniping attacks.
  178. - Have new people be in limbo and need to demonstrate usefulness
  179. before we approve them.
  180. - Clients should estimate their skew as median of skew from servers
  181. over last N seconds.
  182. - Security
  183. - Alices avoid duplicate class C nodes.
  184. - Analyze how bad the partitioning is or isn't.
  185. . Update the hidden service stuff for the new dir approach.
  186. - switch to an ascii format, maybe sexpr?
  187. - authdirservers publish blobs of them.
  188. - other authdirservers fetch these blobs.
  189. - hidserv people have the option of not uploading their blobs.
  190. - you can insert a blob via the controller.
  191. - and there's some amount of backwards compatibility.
  192. - teach clients, intro points, and hidservs about auth mechanisms.
  193. - come up with a few more auth mechanisms.
  194. . Come up with a coherent strategy for bandwidth buckets and TLS. (The
  195. logic for reading from TLS sockets is likely to overrun the bandwidth
  196. buckets under heavy load. (Really, the logic was never right in the
  197. first place.) Also, we should audit all users of get_pending_bytes().)
  198. - Make it harder to circumvent bandwidth caps: look at number of bytes
  199. sent across sockets, not number sent inside TLS stream.
  200. - Make router_is_general_exit() a bit smarter once we're sure what it's for.
  201. - rewrite how libevent does select() on win32 so it's not so very slow.
  202. - Write limiting; separate token bucket for write
  203. - Audit everything to make sure rend and intro points are just as likely to
  204. be us as not.
  205. - Do something to prevent spurious EXTEND cells from making middleman
  206. nodes connect all over. Rate-limit failed connections, perhaps?
  207. Major items for 0.1.2.x:
  208. - Directory guards
  209. R - Server usability
  210. N - Better hidden service performance
  211. - Improve controller
  212. - Asynchronous DNS
  213. - Better estimates in the directory of whether servers have good uptime
  214. (high expected time to failure) or good guard qualities (high
  215. fractional uptime).
  216. - memory usage on dir servers.
  217. copy less!
  218. N - oprofile including kernel time.
  219. Topics to think about during 0.1.2.x development:
  220. - Figure out non-clique.
  221. - Figure out partial network knowledge.
  222. - Figure out incentives.
  223. Future version:
  224. - Limit to 2 dir, 2 OR, N SOCKS connections per IP.
  225. - Handle full buffers without totally borking
  226. - Rate-limit OR and directory connections overall and per-IP and
  227. maybe per subnet.
  228. - Hold-open-until-flushed now works by accident; it should work by
  229. design.
  230. - DoS protection: TLS puzzles, public key ops, bandwidth exhaustion.
  231. - Specify?
  232. - tor-resolve script should use socks5 to get better error messages.
  233. - make min uptime a function of the available choices (say, choose 60th
  234. percentile, not 1 day.)
  235. - Track uptime as %-of-time-up, as well as time-since-last-down.
  236. - hidserv offerers shouldn't need to define a SocksPort
  237. * figure out what breaks for this, and do it.
  238. - auth mechanisms to let hidden service midpoint and responder filter
  239. connection requests.
  240. - Relax clique assumptions.
  241. X start handling server descriptors without a socksport?
  242. - tor should be able to have a pool of outgoing IP addresses
  243. that it is able to rotate through. (maybe)
  244. Blue-sky:
  245. - Patch privoxy and socks protocol to pass strings to the browser.
  246. - Standby/hotswap/redundant hidden services.
  247. - Robust decentralized storage for hidden service descriptors.
  248. - The "China problem"
  249. - Allow small cells and large cells on the same network?
  250. - Cell buffering and resending. This will allow us to handle broken
  251. circuits as long as the endpoints don't break, plus will allow
  252. connection (tls session key) rotation.
  253. - Implement Morphmix, so we can compare its behavior, complexity, etc.
  254. - Other transport. HTTP, udp, rdp, airhook, etc. May have to do our own
  255. link crypto, unless we can bully openssl into it.
  256. - Need a relay teardown cell, separate from one-way ends.
  257. (Pending a user who needs this)
  258. - Handle half-open connections: right now we don't support all TCP
  259. streams, at least according to the protocol. But we handle all that
  260. we've seen in the wild.
  261. (Pending a user who needs this)