4 SPEC - Spec not finalized
16 X <nickm> "Let's try to find a way to make it run and make the version
17 match, but if not, let's just make it run."
18 X <arma> "should we detect if we have a --with-ssl-dir and try the -R
19 by default, if it works?"
21 Items for 0.1.2.x, real soon now:
22 ? - Bug: combination of things:
23 When we've been idle a long time, we stop fetching server
24 descriptors. When we then get a socks request, we build circuits
25 immediately using whatever descriptors we have, rather than waiting
26 until we've fetched correct ones.
28 N - Test guard unreachable logic; make sure that we actually attempt to
29 connect to guards that we think are unreachable from time to time.
30 Make sure that we don't freak out when the network is down.
33 o weight dir requests by advertised bandwidth? with maybe a lower cutoff
34 than for tor traffic. perhaps also weighted by the expected size of
37 . Have (and document) a BEGIN_DIR relay cell that means "Connect to your
40 R - turn the received socks addr:port into a digest for setting .exit
41 - be able to connect without having a server descriptor, to bootstrap.
42 R - handle connect-dir streams that don't have a chosen_exit_name set.
43 o include ORPort in DirServers lines so we can know where to connect.
44 list the orport as 0 if it can't handle begin_dir.
45 o List orports of actual dirservers..
47 - Servers are easy to setup and run: being a relay is about as easy as
49 . Reduce resource load
50 o A way to alert controller when router flags change.
51 o Specify: SETEVENTS NS
53 - Hunt for places that change networkstatus info that I might have
55 - "NS" appears to be used to notify of routerstatus updates.
56 Do we want an actual NS event, for when new networkstatuses
57 arrive, and then rename this one?
59 . option to dl directory info via tor:
60 TunnelDirConns and PreferTunneledDirConns
61 R - actually cause the directory.c functions to know about or_port
62 and use it when we're supposed to.
63 o for tunneled edge conns, stop reading to the bridge connection
64 when the or_conn we're writing to has a full outbuf.
65 o make directory bridge data not get produced when the corresponding
66 or_conn is full, and accept the sometimes directory data will just
71 - Make evdns use windows strerror equivalents.
72 . Make sure patches get into libevent.
73 - Verify that it works well on windows
74 o Debug and re-enable server-side reverse DNS caching
76 - Critical but minor bugs, backport candidates.
77 - support dir 503s better
78 o clients don't log as loudly when they receive them
79 o they don't count toward the 3-strikes rule
80 D But eventually, we give up after getting a lot of 503s.
81 D Delay when we get a lot of 503s, rather than punting onto the
82 servers that have given us 503s?
83 o Add a 'BadDirectory' flag to statuses.
84 o authorities should *never* 503 a cache, and should never 503
85 network status requests.
86 D They can 503 client descriptor requests when they feel like it.
87 How can they distinguish? Not implemented for now, maybe
89 - update dir-spec with what we decided for each of these
91 NR. Write path-spec.txt
94 - Profile client and server; fix slow spots
95 - Address XXX012 items
98 - Tell people about OSX Uninstaller
99 - Quietly document NT Service options
100 - Switch canonical win32 compiler to mingw.
101 - If we haven't replaced privoxy, lock down its configuration in all
102 packages, as documented in tor-doc-unix.html
105 - More prominently, we should have a recommended apps list.
107 - unrecommend IE because of ftp:// bug.
108 N - we should add a preamble to tor-design saying it's out of date.
109 N - Document transport and natdport
111 - Improvements to bandwidth counting
112 R - look into "uncounting" bytes spent on local connections, so
113 we can bandwidthrate but still have fast downloads.
114 R - "bandwidth classes", for incoming vs initiated-here conns,
115 and to give dir conns lower priority.
117 - Forward compatibility fixes
118 - Caches should start trying to cache consensus docs?
120 N - Implement, if we think it's smart.
121 - Start uploading short and long descriptors; authorities should support
122 URLs to retrieve long descriptors, and should discard short descriptors
123 for now. Later, once tools use the "long descriptor" URLs, authorities
124 will serve the short descriptors every time they're asked for
127 N - Implement, if we think it's smart.
128 - Check for any outstanding checks we do on the form or number of client
129 certificates that would prevent us from executing certain
130 blocking-resistance strategies.
133 Topics to think about during 0.1.2.x development:
134 * Figure out incentives.
135 - (How can we make this tolerant of a bad v0?)
136 * Figure out non-clique.
138 - Figure out partial network knowledge.
139 - Figure out hidden services.
140 - Design next-version protocol for directories
141 - Design next-version protocol for connections
143 Deferred from 0.1.2.x:
144 - finish status event implementation and accompanying getinfos
145 - More work on AvoidDiskWrites?
146 NR- Get some kind of "meta signing key" to be used solely to sign
147 releases/to certify releases when signed by the right people/
148 to certify sign the right people's keys? Also use this to cert the SSL
150 - per-conn write buckets
151 - separate config options for read vs write limiting
152 (It's hard to support read > write, since we need better
153 congestion control to avoid overfull buffers there. So,
154 defer the whole thing.)
155 P - Figure out why dll's compiled in mingw don't work right in WinXP.
156 P - Figure out why openssl 0.9.8d "make test" fails at sha256t test.
157 - don't do dns hijacking tests if we're reject *:* exit policy?
158 (deferred until 0.1.1.x is less common)
160 - RAM use in directory authorities.
161 - Memory use improvements:
162 - Look into pulling serverdescs off buffers as they arrive.
163 - Save and mmap v1 directories, and networkstatus docs; store them
164 zipped, not uncompressed.
165 - Switch cached_router_t to use mmap.
166 - What to do about reference counts on windows? (On Unix, this is
167 easy: unlink works fine. (Right?) On Windows, I have doubts. Do we
168 need to keep multiple files?)
169 - What do we do about the fact that people can't read zlib-
170 compressed files manually?
172 - Change the way we handle cells, flow-control, and bridges.
173 - The issue is that we package from edge connections aggressively until
174 we hit their package windows or the circuit package windows, even if
175 the buffer on the corresponding OR connection is pretty damn big. This
176 sucks from a RAM usage POV. Now, we could try to stop reading on the
177 edges (or just the edges connected to a local bridge) when an or_conn's
178 outbuf is full. But if we're a server, and we stop reading on some
179 exit conns when OR conns are full, soon OR conns will contain only
180 traffic from other OR conns, and the exit data in question will never
182 - Remove socketpair-based bridges: use shared (or connected) buffers for
183 communication, rather than sockets.
184 - When relaying cells from an OR conn to an OR conn, have them wait in a
185 queue on the or_circuit_t object; don't move them onto the target conn
186 until there is space in the target conn's outbuf. Also, only package
187 data from exitconns when there is space in the target conn's outbuf.
188 - As an added advantage, this would let us kill stalled _circuits_
189 when their buffers get too full, rather than killing entire OR
190 conns. But we must think about anonymity implications of that.
191 - We'll probably want to do some kind of big refactoring of our
192 dataflow when we do these changes; stuff is hairy enough already,
193 and it will only get harrier with this stuff.
195 - If the client's clock is too far in the past, it will drop (or
196 just not try to get) descriptors, so it'll never build circuits.
197 - Tolerate clock skew on bridge relays.
199 - A more efficient dir protocol.
200 - Authorities should fetch the network-statuses amongst each
201 other, consensus them, and advertise a communal network-status.
202 This is not so much for safety/complexity as it is to reduce
203 bandwidth requirements for Alice.
204 - How does this interact with our goal of being able to choose
205 your own dir authorities? I guess we're now assuming that all
206 dir authorities know all the other authorities in their "group"?
207 - Should we also look into a "delta since last network-status
208 checkpoint" scheme, to reduce overhead further?
209 - Extend the "r" line in network-status to give a set of buckets (say,
210 comma-separated) for that router.
211 - Buckets are deterministic based on IP address.
212 - Then clients can choose a bucket (or set of buckets) to
215 - Improvements to versioning.
216 - When we connect to a Tor server, it sends back a cell listing
217 the IP it believes it is using. Use this to block dvorak's attack.
218 Also, this is a fine time to say what time you think it is.
219 o Verify that a new cell type is okay with deployed codebase
220 . Specify HELLO cells
221 . Figure out v0 compatibility.
224 - Eventdns improvements
225 - Have a way to query for AAAA and A records simultaneously.
226 - Improve request API: At the very least, add the ability to construct
227 a more-or-less arbitrary request and get a response.
228 - (Can we suppress cnames? Should we?)
230 - Now that we're avoiding exits when picking non-exit positions,
231 we need to consider how to pick nodes for internal circuits. If
232 we avoid exits for all positions, we skew the load balancing. If
233 we accept exits for all positions, we leak whether it's an internal
234 circuit at every step. If we accept exits only at the last hop, we
235 reintroduce Lasse's attacks from the Oakland paper.
237 - We should ship with a list of stable dir mirrors -- they're not
238 trusted like the authorities, but they'll provide more robustness
239 and diversity for bootstrapping clients.
241 - Simplify authority operation
242 - Follow weasel's proposal, crossed with mixminion dir config format
244 - A way to adjust router flags from the controller.
245 (How do we prevent the authority from clobbering them soon after?)
246 - a way to pick entry guards based wholly on extend_info equivalent;
247 a way to export extend_info equivalent.
249 - Better estimates in the directory of whether servers have good uptime
250 (high expected time to failure) or good guard qualities (high
252 - AKA Track uptime as %-of-time-up, as well as time-since-last-down
254 - Have a "Faster" status flag that means it. Fast2, Fast4, Fast8?
258 - Failed rend desc fetches sometimes don't get retried. True/false?
260 - Windows server usability
261 - Solve the ENOBUFS problem.
262 - make tor's use of openssl operate on buffers rather than sockets,
263 so we can make use of libevent's buffer paradigm once it has one.
264 - make tor's use of libevent tolerate either the socket or the
265 buffer paradigm; includes unifying the functions in connect.c.
266 - We need a getrlimit equivalent on Windows so we can reserve some
267 file descriptors for saving files, etc. Otherwise we'll trigger
268 asserts when we're out of file descriptors and crash.
269 M - rewrite how libevent does select() on win32 so it's not so very slow.
272 - Add an option (related to AvoidDiskWrites) to disable directory caching.
274 - More status event features:
277 - BAD_DIR_RESPONSE (Unexpected directory response; maybe we're behind
279 - BAD_PROXY (Bad http or https proxy)
280 - UNRECOGNIZED_ROUTER (a nickname we asked for is unavailable)
281 - Status events related to hibernation
282 - something about failing to parse our address?
283 from resolve_my_address() in config.c
284 - sketchy OS, sketchy threading
285 - too many onions queued: threading problems or slow CPU?
287 - TIMEOUT on CHECKING_REACHABILITY
288 - GETINFO status/client, status/server, status/general: There should be
289 some way to learn which status events are currently "in effect."
290 We should specify which these are, what format they appear in, and so
293 Minor items for 0.1.2.x as time permits:
294 - include bandwidth breakdown by conn->type in BW events.
295 - getinfo ns/name/moria2 doesn't include a "v" line, even when some
296 network-statuses I have show it. I suppose the fix should go in
297 networkstatus_getinfo_helper_single() in routerlist.c.
298 - Unify autoconf search code for libevent and openssl. Make code
299 suggest platform-appropriate "devel" / "dev" / whatever packages
300 if we can link but we can't find the headers.
301 - Recommend polipo? Please?
302 - Make documentation realize that location of system configuration file
303 will depend on location of system defaults, and isn't always /etc/torrc.
304 - Review torrc.sample to make it more discursive.
305 o when reporting clock skew (both to logs and to controller), if it's
306 taken 126 seconds to read from the directory, our clock skew estimate
307 is 126 seconds wrong. use conn->timestamp_create or _lastwritten
308 for a closer estimate?
309 - a way to generate the website diagrams from source, so we can
310 translate them as utf-8 text rather than with gimp.
311 R - add d64 and fp64 along-side d and fp so people can paste status
312 entries into a url. since + is a valid base64 char, only allow one
313 at a time. spec and then do.
314 - When we export something from foo.c file for testing purposes only,
315 make a foo_test.h file for test.c to include.
316 - The Debian package now uses --verify-config when (re)starting,
317 to distinguish configuration errors from other errors. Perhaps
318 the RPM and other startup scripts should too?
319 - add a "default.action" file to the tor/vidalia bundle so we can fix the
320 https thing in the default configuration:
321 http://wiki.noreply.org/noreply/TheOnionRouter/TorFAQ#PrivoxyWeirdSSLPort
322 . Flesh out options_description array in src/or/config.c
323 - Don't let 'newnym' be triggered more often than every n seconds.
324 X If we try to publish as a nickname that's already claimed, should
325 we append a number (or increment the number) and try again? This
326 way people who read their logs can fix it as before, but people
327 who don't read their logs will still offer Tor servers.
328 - Fall back to unnamed; warn user; send controller event. ("When we
329 notice a 'Rejected: There is already a named server with this nickname'
330 message... or maybe instead when we see in the networkstatuses that
331 somebody else is Named with the name we want: warn the user, send a
332 STATUS_SERVER message, and fall back to unnamed.")
333 ! - Tor should bind its ports before dropping privs, so users don't
334 have to do the ipchains dance.
335 - Rate limit exit connections to a given destination -- this helps
336 us play nice with websites when Tor users want to crawl them; it
337 also introduces DoS opportunities.
338 - Streamline how we pick entry nodes: Make choose_random_entry() have
339 less magic and less control logic.
340 - Christian Grothoff's attack of infinite-length circuit.
341 the solution is to have a separate 'extend-data' cell type
342 which is used for the first N data cells, and only
343 extend-data cells can be extend requests.
344 - Specify, including thought about anonymity implications.
345 - Display the reasons in 'destroy' and 'truncated' cells under some
347 - We need a way for the authorities to declare that nodes are
348 in a family. Also, it kinda sucks that family declarations use O(N^2)
349 space in the descriptors.
350 - If the server is spewing complaints about raising your ulimit -n,
351 we should add a note about this to the server descriptor so other
352 people can notice too.
354 - see if we should make use of truncate to retry
355 X kill dns workers more slowly
357 . Some back-out mechanism for auto-approval
358 - a way of rolling back approvals to before a timestamp
359 - Consider minion-like fingerprint file/log combination.
360 - packaging and ui stuff:
361 . multiple sample torrc files
362 . figure out how to make nt service stuff work?
364 - Vet all pending installer patches
365 - Win32 installer plus privoxy, sockscap/freecap, etc.
366 - Vet win32 systray helper code
369 - a NEWSTATUS event similar to NEWDESC.
370 - change circuit status events to give more details, like purpose,
371 whether they're internal, when they become dirty, when they become
372 too dirty for further circuits, etc.
373 - What do we want here, exactly?
374 - Specify and implement it.
375 - Change stream status events analogously.
376 - What do we want here, exactly?
377 - Specify and implement it.
378 - Make other events "better".
379 - Change stream status events analogously.
380 - What do we want here, exactly?
381 - Specify and implement it.
382 - Make other events "better" analogously
383 - What do we want here, exactly?
384 - Specify and implement it.
385 . Expose more information via getinfo:
386 - import and export rendezvous descriptors
387 - Review all static fields for additional candidates
388 - Allow EXTENDCIRCUIT to unknown server.
389 - We need some way to adjust server status, and to tell tor not to
390 download directories/network-status, and a way to force a download.
391 - Make everything work with hidden services
392 - Directory system improvements
393 - config option to publish what ports you listen on, beyond
394 ORPort/DirPort. It should support ranges and bit prefixes (?) too.
396 - Relay this in networkstatus.
399 - Need a way to request address lookups (and allocate a stream ID for
400 them) without having a corresponding client socket.
401 - Once this is done, it would be nice to have a way to request address
402 lookups from the controller without using SOCKS.
403 - Refactor exit/entry side of DNS resolve: we don't need a connection_t;
404 we can have an edge_connection_t and (say) dns_request_t both extend an
405 edge_stream_t, and have p_streams and n_streams both be linked lists
408 - Look into generating torrc.{complete|sample}.in, tor.1.in,
409 the HTML manual, and the online config documentation from a single
411 - torrc.complete.in needs attention?
414 - Configuration format really wants sections.
415 - Good RBL substitute.
416 - Authorities should try using exits for http to connect to some URLS
417 (specified in a configuration file, so as not to make the List Of Things
418 Not To Censor completely obvious) and ask them for results. Exits that
419 don't give good answers should have the BadExit flag set.
420 - Our current approach to block attempts to use Tor as a single-hop proxy
421 is pretty lame; we should get a better one.
422 . Update the hidden service stuff for the new dir approach.
423 - switch to an ascii format, maybe sexpr?
424 - authdirservers publish blobs of them.
425 - other authdirservers fetch these blobs.
426 - hidserv people have the option of not uploading their blobs.
427 - you can insert a blob via the controller.
428 - and there's some amount of backwards compatibility.
429 - teach clients, intro points, and hidservs about auth mechanisms.
430 - come up with a few more auth mechanisms.
431 - auth mechanisms to let hidden service midpoint and responder filter
433 - Bind to random port when making outgoing connections to Tor servers,
434 to reduce remote sniping attacks.
435 - Have new people be in limbo and need to demonstrate usefulness
436 before we approve them.
437 - Clients should estimate their skew as median of skew from servers
439 - Make router_is_general_exit() a bit smarter once we're sure what it's for.
440 - Audit everything to make sure rend and intro points are just as likely to
442 - Do something to prevent spurious EXTEND cells from making middleman
443 nodes connect all over. Rate-limit failed connections, perhaps?
444 - Automatically determine what ports are reachable and start using
445 those, if circuits aren't working and it's a pattern we recognize
446 ("port 443 worked once and port 9001 keeps not working").
447 - Limit to 2 dir, 2 OR, N SOCKS connections per IP.
448 - Handle full buffers without totally borking
449 - Rate-limit OR and directory connections overall and per-IP and
451 - Hold-open-until-flushed now works by accident; it should work by
453 - DoS protection: TLS puzzles, public key ops, bandwidth exhaustion.
455 - tor-resolve script should use socks5 to get better error messages.
456 - hidserv offerers shouldn't need to define a SocksPort
457 * figure out what breaks for this, and do it.
458 - tor should be able to have a pool of outgoing IP addresses
459 that it is able to rotate through. (maybe)
460 - Specify; implement.
461 - let each hidden service (or other thing) specify its own
463 - Stop using tor_socketpair to make connection bridges: do an
464 implementation that uses buffers only.
467 - Patch privoxy and socks protocol to pass strings to the browser.
468 - Standby/hotswap/redundant hidden services.
469 - Robust decentralized storage for hidden service descriptors.
470 - The "China problem"
471 - Allow small cells and large cells on the same network?
472 - Cell buffering and resending. This will allow us to handle broken
473 circuits as long as the endpoints don't break, plus will allow
474 connection (tls session key) rotation.
475 - Implement Morphmix, so we can compare its behavior, complexity, etc.
476 - Other transport. HTTP, udp, rdp, airhook, etc. May have to do our own
477 link crypto, unless we can bully openssl into it.
478 - Need a relay teardown cell, separate from one-way ends.
479 (Pending a user who needs this)
480 - Handle half-open connections: right now we don't support all TCP
481 streams, at least according to the protocol. But we handle all that
482 we've seen in the wild.
483 (Pending a user who needs this)
486 - Mark up spec; note unclear points about servers
487 - Mention controller libs someplace.
488 . more pictures from ren. he wants to describe the tor handshake
489 NR- write a spec appendix for 'being nice with tor'
490 - tor-in-the-media page
491 - Remove need for HACKING file.
492 - Figure out licenses for website material.
493 - Specify the keys and key rotation schedules and stuff
496 - and remove home and make the "Tor" picture be the link to home.
497 - put the logo on the website, in source form, so people can put it on
498 stickers directly, etc.
499 R - make a page with the hidden service diagrams.
501 - ask Jan to be the translation coordinator? add to volunteer page.
503 - add a page for localizing all tor's components.
505 - It would be neat if we had a single place that described _all_ the
506 tor-related tools you can use, and what they give you, and how well they
507 work. Right now, we don't give a lot of guidance wrt
508 torbutton/foxproxy/privoxy/polipo in any consistent place.