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.
31 o Stop recommending exits as guards?
32 look at the overall fraction of exits in the network. if the
33 fraction is too small, none of them get to be guards.
35 R - Reconstruct ChangeLog; put rolled-up info in ReleaseNotes or something.
38 - enumerate events of important things that occur in tor, so vidalia can
40 o Backend implementation
41 R - Actually list all the events (notice and warn log messages are a good
42 place to look.) Divide messages into categories, perhaps.
43 R - Specify general event system
44 R - Specify actual events.
45 R - and implement the rest
47 . Have (and document) a BEGIN_DIR relay cell that means "Connect to your
51 o Use for something, so we can be sure it works.
53 R - turn the received socks addr:port into a digest for setting .exit
54 - be able to connect without having a server descriptor, to bootstrap.
55 R - handle connect-dir streams that don't have a chosen_exit_name set.
56 N - include ORPort in DirServers lines so we can know where to connect.
57 list the orport as 0 if it can't handle begin_dir.
58 N - list versions in status page
59 a new line in the status entry. "Tor 0.1.2.2-alpha". If it's
60 a version, treat it like one. If it's something else, assume
61 it's at least 0.1.2.x.
62 maybe we could have it be a new 'v' line in the status, with
63 key=value syntax. so we could have a 'tor' version, but we
64 could also have a 'conn' version, a 'dir' version, etc down
65 the road. and one day maybe the 'tor' key would be deprecated.
67 o Document .noconnect addresses...
68 A new file 'address-spec.txt' that describes .exit, .onion,
71 - Servers are easy to setup and run: being a relay is about as easy as
73 . Reduce resource load
74 o A way to alert controller when router flags change.
75 o Specify: SETEVENTS NS
77 R - Hunt for places that change networkstatus info that I might have
79 R . option to dl directory info via tor
80 o Make an option like __AllDirActionsPrivate that falls back to
81 non-Tor DL when not enough info present. (TunnelDirConns).
82 - Set default to 0 before release candidate.
83 - Think harder about whether TunnelDirConns should be on
85 - Handle case where we have no descriptors and so don't know who can
89 o Don't ask reject *:* nodes for DNS unless client wants you to.
91 - Make evdns use windows strerror equivalents.
92 - Make sure patches get into libevent.
93 - Verify that it works well on windows
94 . Make reverse DNS work.
95 . Add client-side interface
96 o SOCKS interface: specify
97 o SOCKS interface: implement
98 d - Cache answers client-side
99 o Add to Tor-resolve.py
102 o Check for invalid characters in hostnames before trying to resolve
103 them. (This will help catch attempts do to mean things to our DNS
104 server, and bad software that tries to do DNS lookups on whole URLs.)
105 o address_is_invalid_destination() is the right thing to call here
106 (and feel free to make that function smarter)
107 o add a config option to turn it off.
108 - and a man page for that option
109 - Bug 364: notice when all the DNS requests we get back (including a few
110 well-known sites) are all going to the same place.
111 o Bug 363: Warn and die if we can't find a nameserver and we're running a
112 server; don't fall back to 127.0.0.1.
113 ? - maybe re-check dns when we change IP addresses, rather than
115 - Bug 326: Give fewer error messages from nameservers.
116 - Only warn when _all_ nameservers are down; otherwise info.
117 - Increase timeout; what's industry standard?
118 - Alternatively, raise timeout when nameserver dies but comes back
120 - Don't believe that our sole nameserver is dead? or, not until more
121 failures than it would take to think one of several nameservers was
123 - Possibly, don't warn until second retry of a nameserver gets no
125 - warn if all of your nameservers go down and stay down for like
127 R o Take out the '5 second' timeout from the socks detach schedule.
129 - Performance improvements
131 - Critical but minor bugs, backport candidates.
132 - support dir 503s better
133 o clients don't log as loudly when they receive them
134 N - they don't count toward the 3-strikes rule
135 - should there be some threshold of 503's after which we give up?
136 - Delay when we get a lot of 503s?
137 N - split "router is down" from "dirport shouldn't be tried for a while"?
138 We want a field to hold "when did we last get a 503 from this
139 directory server." Probably, it should go in local_routerstatus_t,
140 not in routerinfo_t, since we can try to use servers as directories
141 before we have their descriptors. Possibly, it should also go in
142 trusted_dir_server_t.
143 - authorities should *never* 503 a cache, and should never 503
144 network status requests. They can 503 client descriptor requests
145 when they feel like it.
146 - update dir-spec with what we decided for each of these
149 Nd- Have a mode that doesn't write to disk much, so we can run Tor on
150 flash memory (e.g. Linksys routers or USB keys).
151 o Add AvoidDiskWrites config option.
152 . only write state file when it's "changed"
153 - crank up the numbers if avoiddiskwrites is on.
154 - some things may not want to get written at all.
155 - stop writing identity key / fingerprint / etc every restart
158 NR. Write path-spec.txt
161 - Tell people about OSX Uninstaller
162 - Quietly document NT Service options
163 - Switch canonical win32 compiler to mingw.
164 NR D Get some kind of "meta signing key" to be used solely to sign
165 releases/to certify releases when signed by the right people/
166 to certify sign the right people's keys? Also use this to cert the SSL
168 - If we haven't replaced privoxy, lock down its configuration in all
169 packages, as documented in tor-doc-unix.html
170 o script to look at config.c, torrc.sample, tor.1.in, to tell us
171 what's missing in which and notice which descriptions are missing.
174 - More prominently, we should have a recommended apps list.
176 - unrecommend IE because of ftp:// bug.
177 - torrc.complete.in needs attention?
178 - we should add a preamble to tor-design saying it's out of date.
180 - Improvements to bandwidth counting
181 R - look into "uncounting" bytes spent on local connections, so
182 we can bandwidthrate but still have fast downloads.
183 R - "bandwidth classes", for incoming vs initiated-here conns,
184 and to give dir conns lower priority.
185 . Write limiting; separate token bucket for write
186 o preemptively give a 503 to some v1 dir requests
187 - preemptively give a 503 to some v2 dir requests
188 - per-conn write buckets
189 - separate config options for read vs write limiting
191 Topics to think about during 0.1.2.x development:
192 * Figure out incentives.
193 - (How can we make this tolerant of a bad v0?)
194 * Figure out non-clique.
196 - Figure out partial network knowledge.
197 - Figure out hidden services.
198 - Design next-version protocol for directories
199 - Design next-version protocol for connections
201 For blocking-resistance scheme:
202 o allow ordinary-looking ssl for dir connections. need a new dirport
203 for this, or can we handle both ssl and non-ssl, or should we
204 entirely switch to ssl in certain cases?
205 D need to figure out how to fetch status of a few servers from the BDA
206 without fetching all statuses. A new URL to fetch I presume?
208 Deferred from 0.1.2.x:
209 P - Figure out why dll's compiled in mingw don't work right in WinXP.
210 P - Figure out why openssl 0.9.8d "make test" fails at sha256t test.
212 - RAM use in directory authorities.
213 - Memory use improvements:
214 - Look into pulling serverdescs off buffers as they arrive.
215 - Save and mmap v1 directories, and networkstatus docs; store them
216 zipped, not uncompressed.
217 - Switch cached_router_t to use mmap.
218 - What to do about reference counts on windows? (On Unix, this is
219 easy: unlink works fine. (Right?) On Windows, I have doubts. Do we
220 need to keep multiple files?)
221 - What do we do about the fact that people can't read zlib-
222 compressed files manually?
224 - Refactor DNS resolve implementation
225 - Refactor exit side of resolve: do we need a connection_t?
226 - Refactor entry side of resolve: do we need a connection_t?
228 - If the client's clock is too far in the past, it will drop (or
229 just not try to get) descriptors, so it'll never build circuits.
230 - Tolerate clock skew on bridge relays.
232 - A more efficient dir protocol.
233 - Authorities should fetch the network-statuses amongst each
234 other, consensus them, and advertise a communal network-status.
235 This is not so much for safety/complexity as it is to reduce
236 bandwidth requirements for Alice.
237 - How does this interact with our goal of being able to choose
238 your own dir authorities? I guess we're now assuming that all
239 dir authorities know all the other authorities in their "group"?
240 - Should we also look into a "delta since last network-status
241 checkpoint" scheme, to reduce overhead further?
242 - Extend the "r" line in network-status to give a set of buckets (say,
243 comma-separated) for that router.
244 - Buckets are deterministic based on IP address.
245 - Then clients can choose a bucket (or set of buckets) to
248 - Improvements to versioning.
249 - When we connect to a Tor server, it sends back a cell listing
250 the IP it believes it is using. Use this to block dvorak's attack.
251 Also, this is a fine time to say what time you think it is.
252 o Verify that a new cell type is okay with deployed codebase
253 . Specify HELLO cells
254 . Figure out v0 compatibility.
257 - Eventdns improvements
258 - Have a way to query for AAAA and A records simultaneously.
259 - Improve request API: At the very least, add the ability to construct
260 a more-or-less arbitrary request and get a response.
261 - (Can we suppress cnames? Should we?)
263 - Now that we're avoiding exits when picking non-exit positions,
264 we need to consider how to pick nodes for internal circuits. If
265 we avoid exits for all positions, we skew the load balancing. If
266 we accept exits for all positions, we leak whether it's an internal
267 circuit at every step. If we accept exits only at the last hop, we
268 reintroduce Lasse's attacks from the Oakland paper.
270 - We should ship with a list of stable dir mirrors -- they're not
271 trusted like the authorities, but they'll provide more robustness
272 and diversity for bootstrapping clients.
274 - Simplify authority operation
275 - Follow weasel's proposal, crossed with mixminion dir config format
277 - A way to adjust router flags from the controller.
278 (How do we prevent the authority from clobbering them soon after?)
279 - a way to pick entry guards based wholly on extend_info equivalent;
280 a way to export extend_info equivalent.
282 - Count TLS bandwidth more accurately
284 - Better estimates in the directory of whether servers have good uptime
285 (high expected time to failure) or good guard qualities (high
287 - AKA Track uptime as %-of-time-up, as well as time-since-last-down
289 - Have a "Faster" status flag that means it. Fast2, Fast4, Fast8?
293 - Failed rend desc fetches sometimes don't get retried. True/false?
295 - Windows server usability
296 - Solve the ENOBUFS problem.
297 - make tor's use of openssl operate on buffers rather than sockets,
298 so we can make use of libevent's buffer paradigm once it has one.
299 - make tor's use of libevent tolerate either the socket or the
300 buffer paradigm; includes unifying the functions in connect.c.
301 - We need a getrlimit equivalent on Windows so we can reserve some
302 file descriptors for saving files, etc. Otherwise we'll trigger
303 asserts when we're out of file descriptors and crash.
304 M - rewrite how libevent does select() on win32 so it's not so very slow.
307 - Add an option (related to AvoidDiskWrites) to disable directory caching.
309 Minor items for 0.1.2.x as time permits:
310 R - add d64 and fp64 along-side d and fp so people can paste status
311 entries into a url. since + is a valid base64 char, only allow one
312 at a time. spec and then do.
313 D don't do dns hijacking tests if we're reject *:* exit policy?
314 (deferred until 0.1.1.x is less common)
315 - When we export something from foo.c file for testing purposes only,
316 make a foo_test.h file for test.c to include.
317 - The Debian package now uses --verify-config when (re)starting,
318 to distinguish configuration errors from other errors. Perhaps
319 the RPM and other startup scripts should too?
320 - add a "default.action" file to the tor/vidalia bundle so we can fix the
321 https thing in the default configuration:
322 http://wiki.noreply.org/noreply/TheOnionRouter/TorFAQ#PrivoxyWeirdSSLPort
323 o even if your torrc lists yourself in your myfamily line, don't list it in
325 . Flesh out options_description array in src/or/config.c
326 - Don't let 'newnym' be triggered more often than every n seconds.
327 o change log_fn() to log() on notice/warn/err logs where we can.
328 X If we try to publish as a nickname that's already claimed, should
329 we append a number (or increment the number) and try again? This
330 way people who read their logs can fix it as before, but people
331 who don't read their logs will still offer Tor servers.
332 - Fall back to unnamed; warn user; send controller event. ("When we
333 notice a 'Rejected: There is already a named server with this nickname'
334 message... or maybe instead when we see in the networkstatuses that
335 somebody else is Named with the name we want: warn the user, send a
336 STATUS_SERVER message, and fall back to unnamed.")
337 ! - Tor should bind its ports before dropping privs, so users don't
338 have to do the ipchains dance.
339 - Rate limit exit connections to a given destination -- this helps
340 us play nice with websites when Tor users want to crawl them; it
341 also introduces DoS opportunities.
342 o The bw_accounting file should get merged into the state file.
343 - Streamline how we pick entry nodes: Make choose_random_entry() have
344 less magic and less control logic.
345 - Christian Grothoff's attack of infinite-length circuit.
346 the solution is to have a separate 'extend-data' cell type
347 which is used for the first N data cells, and only
348 extend-data cells can be extend requests.
349 - Specify, including thought about anonymity implications.
350 - Display the reasons in 'destroy' and 'truncated' cells under some
352 - We need a way for the authorities to declare that nodes are
353 in a family. Also, it kinda sucks that family declarations use O(N^2)
354 space in the descriptors.
355 - If the server is spewing complaints about raising your ulimit -n,
356 we should add a note about this to the server descriptor so other
357 people can notice too.
359 - see if we should make use of truncate to retry
360 X kill dns workers more slowly
362 . Some back-out mechanism for auto-approval
363 - a way of rolling back approvals to before a timestamp
364 - Consider minion-like fingerprint file/log combination.
365 - packaging and ui stuff:
366 . multiple sample torrc files
367 . figure out how to make nt service stuff work?
369 - Vet all pending installer patches
370 - Win32 installer plus privoxy, sockscap/freecap, etc.
371 - Vet win32 systray helper code
374 - a NEWSTATUS event similar to NEWDESC.
375 - change circuit status events to give more details, like purpose,
376 whether they're internal, when they become dirty, when they become
377 too dirty for further circuits, etc.
378 - What do we want here, exactly?
379 - Specify and implement it.
380 - Change stream status events analogously.
381 - What do we want here, exactly?
382 - Specify and implement it.
383 - Make other events "better".
384 - Change stream status events analogously.
385 - What do we want here, exactly?
386 - Specify and implement it.
387 - Make other events "better" analogously
388 - What do we want here, exactly?
389 - Specify and implement it.
390 . Expose more information via getinfo:
391 - import and export rendezvous descriptors
392 - Review all static fields for additional candidates
393 - Allow EXTENDCIRCUIT to unknown server.
394 - We need some way to adjust server status, and to tell tor not to
395 download directories/network-status, and a way to force a download.
396 - It would be nice to request address lookups from the controller
398 - Make everything work with hidden services
399 - Directory system improvements
400 - config option to publish what ports you listen on, beyond
401 ORPort/DirPort. It should support ranges and bit prefixes (?) too.
403 - Relay this in networkstatus.
406 - Configuration format really wants sections.
407 - Good RBL substitute.
408 - Authorities should try using exits for http to connect to some URLS
409 (specified in a configuration file, so as not to make the List Of Things
410 Not To Censor completely obvious) and ask them for results. Exits that
411 don't give good answers should have the BadExit flag set.
412 - Our current approach to block attempts to use Tor as a single-hop proxy
413 is pretty lame; we should get a better one.
414 . Update the hidden service stuff for the new dir approach.
415 - switch to an ascii format, maybe sexpr?
416 - authdirservers publish blobs of them.
417 - other authdirservers fetch these blobs.
418 - hidserv people have the option of not uploading their blobs.
419 - you can insert a blob via the controller.
420 - and there's some amount of backwards compatibility.
421 - teach clients, intro points, and hidservs about auth mechanisms.
422 - come up with a few more auth mechanisms.
423 - auth mechanisms to let hidden service midpoint and responder filter
425 - Bind to random port when making outgoing connections to Tor servers,
426 to reduce remote sniping attacks.
427 - Have new people be in limbo and need to demonstrate usefulness
428 before we approve them.
429 - Clients should estimate their skew as median of skew from servers
431 - Make router_is_general_exit() a bit smarter once we're sure what it's for.
432 - Audit everything to make sure rend and intro points are just as likely to
434 - Do something to prevent spurious EXTEND cells from making middleman
435 nodes connect all over. Rate-limit failed connections, perhaps?
436 - Automatically determine what ports are reachable and start using
437 those, if circuits aren't working and it's a pattern we recognize
438 ("port 443 worked once and port 9001 keeps not working").
439 - Limit to 2 dir, 2 OR, N SOCKS connections per IP.
440 - Handle full buffers without totally borking
441 - Rate-limit OR and directory connections overall and per-IP and
443 - Hold-open-until-flushed now works by accident; it should work by
445 - DoS protection: TLS puzzles, public key ops, bandwidth exhaustion.
447 - tor-resolve script should use socks5 to get better error messages.
448 - hidserv offerers shouldn't need to define a SocksPort
449 * figure out what breaks for this, and do it.
450 - tor should be able to have a pool of outgoing IP addresses
451 that it is able to rotate through. (maybe)
452 - Specify; implement.
453 - let each hidden service (or other thing) specify its own
455 - Stop using tor_socketpair to make connection bridges: do an
456 implementation that uses buffers only.
459 - Patch privoxy and socks protocol to pass strings to the browser.
460 - Standby/hotswap/redundant hidden services.
461 - Robust decentralized storage for hidden service descriptors.
462 - The "China problem"
463 - Allow small cells and large cells on the same network?
464 - Cell buffering and resending. This will allow us to handle broken
465 circuits as long as the endpoints don't break, plus will allow
466 connection (tls session key) rotation.
467 - Implement Morphmix, so we can compare its behavior, complexity, etc.
468 - Other transport. HTTP, udp, rdp, airhook, etc. May have to do our own
469 link crypto, unless we can bully openssl into it.
470 - Need a relay teardown cell, separate from one-way ends.
471 (Pending a user who needs this)
472 - Handle half-open connections: right now we don't support all TCP
473 streams, at least according to the protocol. But we handle all that
474 we've seen in the wild.
475 (Pending a user who needs this)
478 - Mark up spec; note unclear points about servers
479 - Mention controller libs someplace.
480 . more pictures from ren. he wants to describe the tor handshake
481 NR- write a spec appendix for 'being nice with tor'
482 - tor-in-the-media page
483 - Remove need for HACKING file.
484 - Figure out licenses for website material.
485 - Specify the keys and key rotation schedules and stuff
488 - and remove home and make the "Tor" picture be the link to home.
489 - put the logo on the website, in source form, so people can put it on
490 stickers directly, etc.
491 R - make a page with the hidden service diagrams.
493 - ask Jan to be the translation coordinator? add to volunteer page.
495 - add a page for localizing all tor's components.