4 SPEC - Spec not finalized
17 Items blocking 0.2.0.10-alpha:
18 - We should back out the MBTF->WFU Guard factors, since they open us
19 up to new attacks, and don't this "median" notion doesn't necessarily
20 help us distinguish between "was good enough to be a guard when
21 we picked it" and "is still adequate to be used as a guard even
22 after we've picked it". We should write a real proposal for this --
24 - Delay the separation of flags till 0.2.1.x. -NM
25 - Let's come up with a good formula for Guard.
29 Take the 7/8s of the Running servers that have the oldest birthdates.
30 To be a guard, you must be in this set and you must at least match the
31 median WFU of the set. In addition, anybody born more than a month ago
32 who has >=50% WFU is always a winner.
34 - Should we ship with a fallback-consensus? Where in the tarball does
35 it go? What's the process for choosing it?
36 - We can, but we don't have to now. Stick it in place of the
37 empty fallback-consensus file in src/config if you like. -NM
38 - To choose, just grab the most recent consensus you have. -NM
40 - If 1.5*MaxCircuitDirtiness is more than KeepAlive, do we then send
41 a KeepAlive and reset our timeout, thus never reaching 1.5*MCD?
42 - Aw, crud. We could keep track of how long it's been since
43 we last did anything _other_ than a keepalive, I guess. -NM
45 o "When reporting clock skew, and we only have a lower bound on
46 the amount of skew, amount anyway, marked as a lower bound.
47 [XXX Nick: what does this mean??]"
50 Things we'd like to do in 0.2.0.x:
51 - See also Flyspray tasks.
52 - See also all items marked XXXX020 and DOCDOC in the code
53 - https://www.torproject.org/eff/legal-faq.html#License doesn't mention
54 licenses for other components of the bundles.
56 - Before the feature freeze: (Nick)
57 D 118 if feasible and obvious
58 D Maintain a skew estimate and use ftime consistently.
59 - 105+TLS, if possible.
60 o Add a separate handshake structure that handles version negotiation,
61 and stores netinfo data until authentication is done.
62 o Revise versions and netinfo to use separate structure; make
63 act-on-netinfo logic separate so it can get called _after_
65 o Variable-length cells
69 o Make VERSIONS variable.
71 o functions to parse x509 certs
72 o functions to validate a single x509 cert against a TLS connection
73 o functions to validate a chain of x509 certs, and extract a PK.
74 o function to encode x509 certs
77 o Keep copies of X509 certs around, not necessarily associated with
81 o Remember certificate digests from TLS
82 o Code to parse and check
84 - Revised handshake: TLS
85 - Server checks for new cipher types, and if it finds them, sends
86 only one cert and does not ask for client certs.
87 - Client sends certs only if server asks for them.
88 - Client sends new cipher list.
89 - Client sends correct extension list.
90 - Revised handshake: post-TLS.
91 o If in 'handshaking' state (since v2+ conn is in use), accept
92 VERSIONS and NETINFO and CERT and LINK_AUTH.
93 o After we send NETINFO, send CERT and LINK_AUTH if needed.
94 - Once we get a good LINK_AUTH, the connection is OPEN.
95 - Ban most cell types on a non-OPEN connection.
96 - Close connections on handshake failure.
97 o Make code work right wrt TLS context rotation.
99 - Don't extend a circuit over a noncanonical connection with
101 - Learn our outgoing IP address from netinfo cells?
103 - Earliest stages of 110 (infinite-length) in v2 protocol:
104 add support for RELAY_EARLY.
105 - Before the feature freeze: (Roger)
106 - Make tunnelled dir conns use begin_dir if enabled
107 - make bridge users fall back from bridge authority to direct attempt
109 - get more v3 authorities before 0.2.0.x comes out.
110 - brainstorm about who those should be
113 - Bug reports Roger has heard along that way that don't have enough
114 details/attention to solve them yet.
115 - tup said that when he set FetchUselessDescriptors, after
116 24 or 48 hours he wasn't fetching any descriptors at all
117 anymore. This was in 0.2.0 but worked fine in 0.1.2.
118 - arma noticed that when his network went away and he tried
119 a new guard node and the connect() syscall failed to it,
120 the guard wasn't being marked as down. 0.2.0.x.
121 - after being without network for 12 hours, arma's tor decided
122 it couldn't fetch any network statuses, and never tried again
123 even when the network came back and arma clicked on things.
125 - phobos says relaybandwidth* sometimes don't do what we expect.
126 http://interloper.org/tmp/2007-06-bw-usage.png
127 I believe this is the result of confusion about whether to use
128 relaybandwidth or bandwidthrate to control bandwidth usage for
129 servers. Chances are, this isn't a bug per se. [phobos]
130 - this notion of authorities notifying servers that they're
131 unreachable is bunk -- it's leftover from the time when all
132 servers ran 24/7. now it triggers every time a server goes
133 away and then returns before the old descriptor has expired.
134 o add a --quiet commandline option that suppresses logs. useful
135 for --hashed-password and maybe others.
136 - Tor logs the libevent version on startup, for debugging purposes.
137 This is great. But it does this before configuring the logs, so
138 it only goes to stdout and is then lost.
139 - we should do another bandwidth test every 12 hours or something
140 if we're showing less than 50KB and our bandwidthrate says we can
141 do more than that. I think some servers are forgetting the results
142 of their first test, and then never seeing use.
145 o 101: Voting on the Tor Directory System (plus 103)
146 D Use if-modified-since on consensus download
148 - GETINFO to get consensus
149 - Event when new consensus arrives
150 - 105: Version negotiation for the Tor protocol
151 . 111: Prioritize local traffic over relayed.
152 - Merge into tor-spec.txt.
155 . Make cells get buffered on circuit, not on the or_conn.
156 . Switch to pool-allocation for cells?
157 - Benchmark pool-allocation vs straightforward malloc.
158 - Adjust memory allocation logic in pools to favor a little less
160 . Remove socketpair-based bridges conns, and the word "bridge". (Use
161 shared (or connected) buffers for communication, rather than sockets.)
163 - Handle rate-limiting on directory writes to linked directory
164 connections in a more sensible manner.
165 - Find more ways to test this.
166 D Do TLS connection rotation more often than "once a week" in the
168 D Streamline how we pick entry nodes: Make choose_random_entry() have
169 less magic and less control logic.
170 - Refactor networkstatus generation:
171 - Include "v" line in getinfo values.
174 . Bridges users (rudimentary version)
175 o Ability to specify bridges manually
176 o Config option 'UseBridges' that bridge users can turn on.
177 o uses bridges as first hop rather than entry guards.
178 o if you don't have any routerinfos for your bridges, or you don't
179 like the ones you have, ask a new bridge for its server/authority.
180 . Ask all directory questions to bridge via BEGIN_DIR.
181 - use the bridges for dir fetches even when our dirport is open.
182 R - drop 'authority' queries if they're to our own identity key; accept
184 X Design/implement the "local-status" or something like it, from the
185 "Descriptor purposes: how to tell them apart" section of
186 http://archives.seul.org/or/dev/May-2007/msg00008.html
187 o timeout and retry schedules for fetching bridge descriptors
188 - give extend_info_t a router_purpose again
189 o react faster to download networkstatuses after the first bridge
191 o be more robust to bridges being marked as down and leaving us
192 stranded without any known "running" bridges.
193 - Bridges operators (rudimentary version)
194 - Ability to act as dir cache without a dir port.
195 o Bridges publish to bridge authorities
196 o Fix BEGIN_DIR so that you connect to bridge of which you only
197 know IP (and optionally fingerprint), and then use BEGIN_DIR to learn
199 - look at server_mode() and decide if it always applies to bridges too.
200 - Bridges authorities (rudimentary version)
201 o Rudimentary "do not publish networkstatus" option for bridge
203 - Clients can ask bridge authorities for more bridges.
205 o Clients can ask bridge authorities for updates on known bridges.
206 - More TLS normalization work: make Tor less easily
208 - Directory system improvements
210 - Make BEGIN_DIR mandatory for asking questions of bridge authorities?
212 - Features (other than bridges):
213 - Audit how much RAM we're using for buffers and cell pools; try to
215 - Base relative control socket paths on datadir.
216 - Make TrackHostExits expire TrackHostExitsExpire seconds after their
217 *last* use, not their *first* use.
218 - Limit to 2 dir, 2 OR, N SOCKS connections per IP.
219 - Or maybe close connections from same IP when we get a lot from one.
220 - Or maybe block IPs that connect too many times at once.
221 D add an AuthDirBadexit torrc option if we decide we want one.
224 N - Hack up a client that gives out weird/no certificates, so we can
225 test to make sure that this doesn't cause servers to crash.
228 - can we deprecate 'getinfo network-status'?
229 - can we deprecate the FastFirstHopPK config option?
233 - Quietly document NT Service options
234 - More prominently, we should have a recommended apps list.
235 - recommend pidgin (gaim is renamed)
236 - unrecommend IE because of ftp:// bug.
237 - we should add a preamble to tor-design saying it's out of date.
238 . Document transport and natdport in a good HOWTO.
239 - Publicize torel. (What else?
240 . Finish path-spec.txt
243 P - Can we switch to polipo?
244 P - Make documentation realize that location of system configuration file
245 will depend on location of system defaults, and isn't always /etc/torrc.
246 P - Figure out why dll's compiled in mingw don't work right in WinXP.
247 - Create packages for Nokia 800, requested by Chris Soghoian
248 P - Consider creating special Tor-Polipo-Vidalia test packages,
249 requested by Dmitri Vitalev
250 o Get Vidalia supporting protocolinfo and using auth by default.
251 P - create a "make win32-bundle" for vidalia-privoxy-tor-torbutton
254 Nice-to-have items for 0.2.0.x, time permitting:
256 - we try to build 4 test circuits to break them over different
257 servers. but sometimes our entry node is the same for multiple
258 test circuits. this defeats the point.
261 - switch out privoxy in the bundles and replace it with polipo.
262 - make the new tls handshake blocking-resistant.
263 - figure out some way to collect feedback about what countries are using
264 bridges, in a way that doesn't screw anonymity too much.
265 - let tor dir mirrors proxy connections to the tor download site, so
266 if you know a bridge you can fetch the tor software.
267 - more strategies for distributing bridge addresses in a way that
268 doesn't rely on knowing somebody who runs a bridge for you.
269 - A way to adjust router status flags from the controller. (How do we
270 prevent the authority from clobbering them soon afterward?)
271 - Bridge authorities should do reachability testing but only on the
272 purpose==bridge descriptors they have.
273 - Clients should estimate their skew as median of skew from servers
275 - Investigate RAM use in Tor servers.
276 - Start on the WSAENOBUFS solution.
277 - Start on Windows auto-update for Tor
279 Deferred from 0.2.0.x:
281 - 113: Simplifying directory authority administration
282 - 110: prevent infinite-length circuits (phase one)
283 - 118: Listen on and advertise multiple ports:
284 - Tor should be able to have a pool of outgoing IP addresses that it is
285 able to rotate through. (maybe. Possible overlap with proposal 118.)
286 - config option to publish what ports you listen on, beyond
287 ORPort/DirPort. It should support ranges and bit prefixes (?) too.
288 (This is very similar to proposal 118.)
290 - Internal code support for ipv6:
291 o Clone ipv6 functions (inet_ntop, inet_pton) where they don't exist.
292 - Most address variables need to become tor_addr_t
293 - Teach resolving code how to handle ipv6.
294 - Teach exit policies about ipv6 (consider ipv4/ipv6 interaction!)
296 - Let controller set router flags for authority to transmit, and for
298 - add an 'exit-address' line in the descriptor for servers that exit
299 from something that isn't their published address.
300 - More work on AvoidDiskWrites?
304 - MAYBE kill stalled circuits rather than stalled connections. This is
305 possible thanks to cell queues, but we need to consider the anonymity
307 - Implement TLS shutdown properly when possible.
309 - If the client's clock is too far in the past, it will drop (or just not
310 try to get) descriptors, so it'll never build circuits.
312 - Make resolves no longer use edge_connection_t unless they are actually
313 _on_ a socks connection: have edge_connection_t and (say)
314 dns_request_t both extend an edge_stream_t, and have p_streams and
315 n_streams both be linked lists of edge_stream_t.
316 - Generate torrc.{complete|sample}.in, tor.1.in, the HTML manual, and the
317 online config documentation from a single source.
318 - Move all status info out of routerinfo into local_routerstatus. Make
319 "who can change what" in local_routerstatus explicit. Make
320 local_routerstatus (or equivalent) subsume all places to go for "what
322 - Blocking/scanning-resistance
323 - It would be potentially helpful to https requests on the OR port by
324 acting like an HTTPS server.
325 - Do we want to maintain our own set of entryguards that we use as
326 next hop after the bridge? Open research question; let's say no
327 for 0.2.0 unless we learn otherwise.
328 - Some mechanism for specifying that we want to stop using a cached
331 - Detect correct version of libraries from autoconf script.
335 - See also Flyspray tasks.
336 - See also all OPEN/ACCEPTED proposals.
337 - See also all items marked XXXX and FFFF in the code.
340 - Our current approach to block attempts to use Tor as a single-hop proxy
341 is pretty lame; we should get a better one.
342 - Allow small cells and large cells on the same network?
343 - Cell buffering and resending. This will allow us to handle broken
344 circuits as long as the endpoints don't break, plus will allow
345 connection (tls session key) rotation.
346 - Implement Morphmix, so we can compare its behavior, complexity,
347 etc. But see paper breaking morphmix.
348 - Other transport. HTTP, udp, rdp, airhook, etc. May have to do our own
349 link crypto, unless we can bully DTLS into it.
350 - Need a relay teardown cell, separate from one-way ends.
351 (Pending a user who needs this)
352 - Handle half-open connections: right now we don't support all TCP
353 streams, at least according to the protocol. But we handle all that
354 we've seen in the wild.
355 (Pending a user who needs this)
359 X turn the received socks addr:port into a digest for setting .exit
360 - handle connect-dir streams that don't have a chosen_exit_name set.
361 - Have a "Faster" status flag that means it. Fast2, Fast4, Fast8?
362 - Add an option (related to AvoidDiskWrites) to disable directory
363 caching. (Is this actually a good idea??)
364 - Add d64 and fp64 along-side d and fp so people can paste status
365 entries into a url. since + is a valid base64 char, only allow one
366 at a time. Consider adding to controller as well.
367 - Some back-out mechanism for auto-approval on authorities
368 - a way of rolling back approvals to before a timestamp
369 - Consider minion-like fingerprint file/log combination.
370 - Have new people be in limbo and need to demonstrate usefulness
371 before we approve them.
374 - Standby/hotswap/redundant hidden services.
375 . Update the hidden service stuff for the new dir approach. (Much
376 of this will be superseded by 114.)
377 - switch to an ascii format, maybe sexpr?
378 - authdirservers publish blobs of them.
379 - other authdirservers fetch these blobs.
380 - hidserv people have the option of not uploading their blobs.
381 - you can insert a blob via the controller.
382 - and there's some amount of backwards compatibility.
383 - teach clients, intro points, and hidservs about auth mechanisms.
384 - come up with a few more auth mechanisms.
385 - auth mechanisms to let hidden service midpoint and responder filter
387 - Let each hidden service (or other thing) specify its own
389 - Hidserv offerers shouldn't need to define a SocksPort
392 X When we notice a 'Rejected: There is already a named server with
393 this nickname' message... or maybe instead when we see in the
394 networkstatuses that somebody else is Named with the name we
395 want: warn the user, send a STATUS_SERVER message, and fall back
397 - If the server is spewing complaints about raising your ulimit -n,
398 we should add a note about this to the server descriptor so other
399 people can notice too.
400 - When we hit a funny error from a dir request (eg 403 forbidden),
401 but tor is working and happy otherwise, and we haven't seen many
402 such errors recently, then don't warn about it.
405 - Implement missing status events and accompanying getinfos
407 - BAD_DIR_RESPONSE (Unexpected directory response; maybe we're behind
409 - BAD_PROXY (Bad http or https proxy)
410 - UNRECOGNIZED_ROUTER (a nickname we asked for is unavailable)
411 - Status events related to hibernation
412 - something about failing to parse our address?
413 from resolve_my_address() in config.c
414 - sketchy OS, sketchy threading
415 - too many onions queued: threading problems or slow CPU?
416 - Implement missing status event fields:
417 - TIMEOUT on CHECKING_REACHABILITY
418 - GETINFO status/client, status/server, status/general: There should be
419 some way to learn which status events are currently "in effect."
420 We should specify which these are, what format they appear in, and so
422 - More information in events:
423 - Include bandwidth breakdown by conn->type in BW events.
424 - Change circuit status events to give more details, like purpose,
425 whether they're internal, when they become dirty, when they become
426 too dirty for further circuits, etc.
427 - Change stream status events analogously.
428 - Expose more information via getinfo:
429 - import and export rendezvous descriptors
430 - Review all static fields for additional candidates
431 - Allow EXTENDCIRCUIT to unknown server.
432 - We need some way to adjust server status, and to tell tor not to
433 download directories/network-status, and a way to force a download.
434 - Make everything work with hidden services
436 - Performance/resources
437 - per-conn write buckets
438 - separate config options for read vs write limiting
439 (It's hard to support read > write, since we need better
440 congestion control to avoid overfull buffers there. So,
441 defer the whole thing.)
442 - Look into pulling serverdescs off buffers as they arrive.
443 - Rate limit exit connections to a given destination -- this helps
444 us play nice with websites when Tor users want to crawl them; it
445 also introduces DoS opportunities.
446 - Consider truncating rather than destroying failed circuits,
447 in order to save the effort of restarting. There are security
448 issues here that need thinking, though.
449 - Handle full buffers without totally borking
450 - Rate-limit OR and directory connections overall and per-IP and
454 - Hold-open-until-flushed now works by accident; it should work by
456 - Display the reasons in 'destroy' and 'truncated' cells under
458 - Make router_is_general_exit() a bit smarter once we're sure what
460 - Automatically determine what ports are reachable and start using
461 those, if circuits aren't working and it's a pattern we
462 recognize ("port 443 worked once and port 9001 keeps not
466 - some better fix for bug #516?
467 - don't do dns hijacking tests if we're reject *:* exit policy?
468 (deferred until 0.1.1.x is less common)
471 - Servers might check certs for known-good ssl websites, and if
472 they come back self-signed, declare themselves to be
473 non-exits. Similar to how we test for broken/evil dns now.
474 - Authorities should try using exits for http to connect to some
475 URLS (specified in a configuration file, so as not to make the
476 List Of Things Not To Censor completely obvious) and ask them
477 for results. Exits that don't give good answers should have
478 the BadExit flag set.
479 - Alternatively, authorities should be able to import opinions
480 from Snakes on a Tor.
481 - More consistent error checking in router_parse_entry_from_string().
482 I can say "banana" as my bandwidthcapacity, and it won't even squeak.
483 - Bind to random port when making outgoing connections to Tor servers,
484 to reduce remote sniping attacks.
485 - Audit everything to make sure rend and intro points are just as
486 likely to be us as not.
487 - Do something to prevent spurious EXTEND cells from making
488 middleman nodes connect all over. Rate-limit failed
489 connections, perhaps?
490 - DoS protection: TLS puzzles, public key ops, bandwidth exhaustion.
493 - Now that we're avoiding exits when picking non-exit positions,
494 we need to consider how to pick nodes for internal circuits. If
495 we avoid exits for all positions, we skew the load balancing. If
496 we accept exits for all positions, we leak whether it's an
497 internal circuit at every step. If we accept exits only at the
498 last hop, we reintroduce Lasse's attacks from the Oakland paper.
500 - Windows server usability
501 - Solve the ENOBUFS problem.
502 - make tor's use of openssl operate on buffers rather than sockets,
503 so we can make use of libevent's buffer paradigm once it has one.
504 - make tor's use of libevent tolerate either the socket or the
505 buffer paradigm; includes unifying the functions in connect.c.
506 - We need a getrlimit equivalent on Windows so we can reserve some
507 file descriptors for saving files, etc. Otherwise we'll trigger
508 asserts when we're out of file descriptors and crash.
509 - Merge code from Urz into libevent
510 - Make Tor use evbuffers.
513 - a way to generate the website diagrams from source, so we can
514 translate them as utf-8 text rather than with gimp. (svg? or
516 . Flesh out options_description array in src/or/config.c
517 . multiple sample torrc files
518 . figure out how to make nt service stuff work?
520 - Refactor tor man page to divide generally useful options from
522 - Add a doxygen style checker to make check-spaces so nick doesn't drift
523 too far from arma's undocumented styleguide. Also, document that
524 styleguide in HACKING. (See r9634 for example.)
525 - exactly one space at beginning and at end of comments, except i
526 guess when there's line-length pressure.
527 - if we refer to a function name, put a () after it.
528 - only write <b>foo</b> when foo is an argument to this function.
529 - doxygen comments must always end in some form of punctuation.
530 - capitalize the first sentence in the doxygen comment, except
532 - avoid spelling errors and incorrect comments. ;)
535 - The Debian package now uses --verify-config when (re)starting,
536 to distinguish configuration errors from other errors. Perhaps
537 the RPM and other startup scripts should too?
538 - add a "default.action" file to the tor/vidalia bundle so we can
539 fix the https thing in the default configuration:
540 http://wiki.noreply.org/noreply/TheOnionRouter/TorFAQ#PrivoxyWeirdSSLPort
543 - Patch privoxy and socks protocol to pass strings to the browser.
546 Documentation, non-version-specific.
548 - Mark up spec; note unclear points about servers
549 NR - write a spec appendix for 'being nice with tor'
550 - Specify the keys and key rotation schedules and stuff
551 - Mention controller libs someplace.
552 - Remove need for HACKING file.
553 P - document http://wiki.noreply.org/noreply/TheOnionRouter/TransparentProxy on freebsd and osx
554 P - figure out why x86_64 won't build rpms from tor.spec
555 P - figure out spec files for bundles of vidalia-tor-polipo
556 P - figure out polipo install scripts for bundles of vidalia-tor-polipo on osx, win32
557 P - figure out selinux policy for tor
558 P - change packaging system to more automated and specific for each
559 platform, suggested by Paul Wouter
560 P - Setup repos for redhat and suse rpms & start signing the rpms the
561 way package management apps prefer
564 J - tor-in-the-media page
565 P - Figure out licenses for website material.
566 (Phobos reccomends the Open Publication License at
567 http://opencontent.org/openpub/)
568 P - put the logo on the website, in source form, so people can put it on
569 stickers directly, etc.
570 P - put the source image for the stickers on the website, so people can
572 P - figure out a license for the logos and docs we publish (trademark
574 (Phobos reccomends the Open Publication License at
575 http://opencontent.org/openpub/)
576 R - make a page with the hidden service diagrams.
577 P - ask Jan/Jens to be the translation coordinator? add to volunteer page.
578 - add a page for localizing all tor's components.
579 - It would be neat if we had a single place that described _all_ the
580 tor-related tools you can use, and what they give you, and how well they
581 work. Right now, we don't give a lot of guidance wrt
582 torbutton/foxproxy/privoxy/polipo in any consistent place.
585 - make a mailing list with the mirror operators
586 - make an automated tool to check /project/trace/ at mirrors to
587 learn which ones are lagging behind.
588 - auto (or manually) cull the mirrors that are broken; and
589 contact their operator?
590 - a set of instructions for mirror operators to make their apaches
591 serve our charsets correctly, and bonus points for language
593 - figure out how to load-balance the downloads across mirrors?
594 - ponder how to get users to learn that they should google for
595 "tor mirrors" if the main site is blocked.
596 - find a mirror volunteer to coordinate all of this