TODO 8.6 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185
  1. Legend:
  2. SPEC!! - Not specified
  3. SPEC - Spec not finalized
  4. NICK - nick claims
  5. ARMA - arma claims
  6. - Not done
  7. * Top priority
  8. . Partially done
  9. o Done
  10. D Deferred
  11. X Abandoned
  12. . streams / circuits
  13. o Implement streams
  14. o Rotate circuits after N minutes?
  15. X Circuits should expire when circuit->expire triggers
  16. NICK . Handle half-open connections
  17. o Figure out what causes connections to close, standardize
  18. when we mark a connection vs when we tear it down
  19. o Look at what ssl does to keep from mutating data streams
  20. ARMA - Reduce streamid footprint from 7 bytes to 2 bytes
  21. - Check for collisions in streamid (now possible with
  22. just 2 bytes), and back up & replace with padding if so
  23. - Use the 3 saved bytes to put pseudorandomness in each relay cell
  24. - Use the 4 reserved bytes in each cell header to keep 1/5
  25. of a sha1 of the relay payload (move into stream header)
  26. - Move length into the stream header too
  27. - Spec the stream_id stuff. Clarify that nobody on the backward
  28. stream should look at stream_id.
  29. ARMA . Exit policies
  30. o Spec how to write the exit policies
  31. - Path selection algorithms
  32. - Let user request certain nodes
  33. - And disallow certain nodes
  34. D Choose path by jurisdiction, etc?
  35. - Make relay end cells have failure status and payload attached
  36. - Streams that fail due to exit policy must reextend to new node
  37. - Add extend_wait state to edge connections, thumb through them
  38. when the AP get an extended cell.
  39. SPEC!! D Non-clique topologies
  40. D Implement our own memory management, at least for common structs
  41. . Appropriate logging
  42. - Come up with convention for what log level means what
  43. - Make code follow convention
  44. . Put CPU workers in separate processes
  45. o Handle multiple cpu workers (one for each cpu, plus one)
  46. o Queue for pending tasks if all workers full
  47. o Support the 'process this onion' task
  48. NICK - Support the 'decrypt this RSA blob' handshake1 task
  49. - Merge dnsworkers and cpuworkers to some extent
  50. - Handle cpuworkers dying
  51. D Support later handshake parts
  52. o Simple directory servers
  53. o Include key in source; sign directories
  54. o Signed directory backend
  55. o Document
  56. o Integrate
  57. o Add versions to code
  58. o Have directories list recommended-versions
  59. o Include (unused) line in directories
  60. o Check for presence of line.
  61. o Quit if running the wrong version
  62. o Command-line option to override quit
  63. o Add more information to directory server entries
  64. o Exit policies
  65. - More directory servers
  66. - Add in long-term nicknames
  67. - Give normal routers signing keys
  68. - Let dirservers keep only {nickname, signingkey} in routers.or
  69. - dirport needs to accept 'post' requests
  70. for routers submitting (signed) new entries
  71. - routers submit new entries periodically
  72. - dirserver checks signature
  73. D client checks signature?
  74. D Advanced directory servers
  75. D Automated reputation management
  76. SPEC!! D Figure out how to do threshold directory servers
  77. D jurisdiction info in dirserver entries? other info?
  78. . Scrubbing proxies
  79. - Find an smtp proxy?
  80. - Check the old smtp proxy code
  81. o Find an ftp proxy? wget --passive
  82. D Wait until there are packet redirectors for Linux
  83. . Get socks4a support into Mozilla
  84. . Get tor to act like a socks server
  85. o socks4, socks4a
  86. D socks5
  87. SPEC!! - Handle socks commands other than connect, eg, bind?
  88. . Develop rendezvous points
  89. o Design
  90. - Spec
  91. - Implement
  92. D Deploy and manage open source development site.
  93. . Documentation
  94. o Discussion of socks, tsocks, etc
  95. o On-the-network protocol
  96. o Onions
  97. o Cells
  98. . Better comments for functions!
  99. - Tests
  100. o Testing harness/infrastructure
  101. NICK . Unit tests
  102. D System tests (how?)
  103. - Performance tests, so we know when we've improved
  104. . webload infrastructure (Bruce)
  105. . httperf infrastructure (easy to set up)
  106. . oprofile (installed in RH >8.0)
  107. D Deploy a widespread network
  108. D Load balancing between router twins
  109. D Keep track of load over links/nodes, to
  110. know who's hosed
  111. NICK . Daemonize and package
  112. o Teach it to fork and background
  113. - Red Hat spec file
  114. - Debian spec file equivalent
  115. . Autoconf
  116. . Which .h files are we actually using?
  117. . Port to:
  118. o Linux
  119. o BSD
  120. . Solaris
  121. o Cygwin
  122. o Win32
  123. o OS X
  124. o openssl randomness
  125. o inet_ntoa
  126. . stdint.h
  127. - Make a script to set up a local network on your machine
  128. D Move away from openssl
  129. o Abstract out crypto calls
  130. D Look at nss, others? Just include code?
  131. o Clearer bandwidth management
  132. o Do we want to remove bandwidth from OR handshakes?
  133. o What about OP handshakes?
  134. - More flexibility in node addressing
  135. D Support IPv6 rather than just 4
  136. - Handle multihomed servers (config variable to set IP)
  137. . Move from onions to ephemeral DH
  138. o incremental path building
  139. o transition circuit-level sendmes to hop-level sendmes
  140. o implement truncate, truncated
  141. o move from 192byte DH to 128byte DH, so it isn't so damn slow
  142. - exiting from not-last hop
  143. - OP logic to decide to extend/truncate a path
  144. - make sure exiting from the not-last hop works
  145. - logic to find last *open* hop, not last hop, in cpath
  146. - choose exit nodes by exit policies
  147. Older (done) todo stuff:
  148. o Use a stronger cipher
  149. o aes now, by including the code ourselves
  150. X On the fly compression of each stream
  151. o Clean up the event loop (optimize and sanitize)
  152. o Remove that awful concept of 'roles'
  153. o Terminology
  154. o Circuits, topics, cells stay named that
  155. o 'Connection' gets divided, or renamed, or something?
  156. o DNS farm
  157. o Distribute queries onto the farm, get answers
  158. o Preemptively grow a new worker before he's needed
  159. o Prune workers when too many are idle
  160. o DNS cache
  161. o Clear DNS cache over time
  162. D Honor DNS TTL info (how??)
  163. o Have strategy when all workers are busy
  164. o Keep track of which connections are in dns_wait
  165. o Need to cache positives/negatives on the tor side
  166. o Keep track of which queries have been asked
  167. o Better error handling when
  168. o An address doesn't resolve
  169. o We have max workers running
  170. o Consider taking the master out of the loop?
  171. D Implement reply onions
  172. o Total rate limiting
  173. o Look at OR handshake in more detail
  174. o Spec it
  175. o Merge OR and OP handshakes
  176. o rearrange connection_or so it doesn't suck so much to read
  177. D Periodic link key rotation. Spec?
  178. o wrap malloc with something that explodes when it fails
  179. o Clean up the number of places that get to look at prkey