1 Changes that will be in 0.0.7pre1:
3 - Fix an epipe bug: sometimes when directory connections failed
4 to connect, we would give them a chance to flush before closing
6 - Directory connections were asking the wrong poll socket to
7 start writing, and not asking themselves to start writing.
8 - When we detached from a circuit because we sent a begin but
9 didn't get a connected, we would use it again the first time;
10 but after that we would correctly switch to a different one.
11 - When we detached from a circuit because of resolvefailed, we
12 would immediately try the same circuit twice more, and then
13 give up on the resolve thinking we'd tried three different
15 - Stop warning when the first onion decrypt attempt fails; they
16 will sometimes legitimately fail now that we rotate keys.
18 - Doxygen markup on all functions.
19 - Rearranged functions more clearly; there are several new source
22 Changes in version 0.0.6.2 - 2004-05-16
23 o Our integrity-checking digest was checking only the most recent cell,
24 not the previous cells like we'd thought.
25 Thanks to Stefan Mark for finding the flaw!
28 Changes in version 0.0.6.1 - 2004-05-06
29 o Fix two bugs in our AES counter-mode implementation (this affected
30 onion-level stream encryption, but not TLS-level). It turns
31 out we were doing something much more akin to a 16-character
32 polyalphabetic cipher. Oops.
33 Thanks to Stefan Mark for finding the flaw!
34 o Retire moria3 as a directory server, and add tor26 as a directory
38 Changes in version 0.0.6 - 2004-05-02
40 - Hidden services and rendezvous points are implemented. Go to
41 http://6sxoyfb3h2nvok2d.onion/ for an index of currently available
42 hidden services. (This only works via a socks4a proxy such as
43 Privoxy, and currently it's quite slow.)
44 - We now rotate link (tls context) keys and onion keys.
45 - CREATE cells now include oaep padding, so you can tell
46 if you decrypted them correctly.
47 - Retry stream correctly when we fail to connect because of
48 exit-policy-reject (should try another) or can't-resolve-address.
49 - When we hup a dirserver and we've *removed* a server from the
50 approved-routers list, now we remove that server from the
51 in-memory directories too.
52 - Add bandwidthburst to server descriptor.
53 - Directories now say which dirserver signed them.
54 - Use a tor_assert macro that logs failed assertions too.
55 - Since we don't support truncateds much, don't bother sending them;
57 - Fetch randomness from /dev/urandom better (not via fopen/fread)
58 - Better debugging for tls errors
59 - Set Content-Type on the directory and hidserv descriptor.
60 - Remove IVs from cipher code, since AES-ctr has none.
62 - Fix an assert trigger for exit nodes that's been plaguing us since
63 the days of 0.0.2prexx (thanks weasel!)
64 - Fix a bug where we were closing tls connections intermittently.
65 It turns out openssl keeps its errors around -- so if an error
66 happens, and you don't ask about it, and then another openssl
67 operation happens and succeeds, and you ask if there was an error,
68 it tells you about the first error.
69 - Fix a bug that's been lurking since 27 may 03 (!)
70 When passing back a destroy cell, we would use the wrong circ id.
71 - Don't crash if a conn that sent a begin has suddenly lost its circuit.
72 - Some versions of openssl have an SSL_pending function that erroneously
73 returns bytes when there is a non-application record pending.
74 - Win32 fixes. Tor now compiles on win32 with no warnings/errors.
75 o We were using an array of length zero in a few places.
76 o Win32's gethostbyname can't resolve an IP to an IP.
77 o Win32's close can't close a socket.
78 o Handle windows socket errors correctly.
80 - check for <sys/limits.h> so we build on FreeBSD again, and
81 <machine/limits.h> for NetBSD.
84 Changes in version 0.0.5 - 2004-03-30
85 o Install torrc as torrc.sample -- we no longer clobber your
87 o Fix mangled-state bug in directory fetching (was causing sigpipes).
88 o Only build circuits after we've fetched the directory: clients were
89 using only the directory servers before they'd fetched a directory.
90 This also means longer startup time; so it goes.
91 o Fix an assert trigger where an OP would fail to handshake, and we'd
92 expect it to have a nickname.
93 o Work around a tsocks bug: do a socks reject when AP connection dies
94 early, else tsocks goes into an infinite loop.
95 o Hold socks connection open until reply is flushed (if possible)
96 o Make exit nodes resolve IPs to IPs immediately, rather than asking
97 the dns farm to do it.
98 o Fix c99 aliasing warnings in rephist.c
99 o Don't include server descriptors that are older than 24 hours in the
101 o Give socks 'reject' replies their whole 15s to attempt to flush,
102 rather than seeing the 60s timeout and assuming the flush had failed.
103 o Clean automake droppings from the cvs repository
104 o Add in a 'notice' log level for things the operator should hear
105 but that aren't warnings
108 Changes in version 0.0.4 - 2004-03-26
109 o When connecting to a dirserver or OR and the network is down,
113 Changes in version 0.0.3 - 2004-03-26
114 o Warn and fail if server chose a nickname with illegal characters
115 o Port to Solaris and Sparc:
116 - include missing header fcntl.h
117 - have autoconf find -lsocket -lnsl automatically
118 - deal with hardware word alignment
119 - make uname() work (solaris has a different return convention)
120 - switch from using signal() to sigaction()
121 o Preliminary work on reputation system:
122 - Keep statistics on success/fail of connect attempts; they're published
123 by kill -USR1 currently.
124 - Add a RunTesting option to try to learn link state by creating test
125 circuits, even when SocksPort is off.
126 - Remove unused open circuits when there are too many.
129 Changes in version 0.0.2 - 2004-03-19
130 - Include strlcpy and strlcat for safer string ops
131 - define INADDR_NONE so we compile (but still not run) on solaris
134 Changes in version 0.0.2pre27 - 2004-03-14
136 - Allow internal tor networks (we were rejecting internal IPs,
137 now we allow them if they're set explicitly).
138 - And fix a few endian issues.
141 Changes in version 0.0.2pre26 - 2004-03-14
143 - If a stream times out after 15s without a connected cell, don't
144 try that circuit again: try a new one.
145 - Retry streams at most 4 times. Then give up.
146 - When a dirserver gets a descriptor from an unknown router, it
147 logs its fingerprint (so the dirserver operator can choose to
148 accept it even without mail from the server operator).
149 - Inform unapproved servers when we reject their descriptors.
150 - Make tor build on Windows again. It works as a client, who knows
152 - Clearer instructions in the torrc for how to set up a server.
153 - Be more efficient about reading fd's when our global token bucket
154 (used for rate limiting) becomes empty.
156 - Stop asserting that computers always go forward in time. It's
158 - When we sent a cell (e.g. destroy) and then marked an OR connection
159 expired, we might close it before finishing a flush if the other
160 side isn't reading right then.
161 - Don't allow dirservers to start if they haven't defined
163 - We were caching transient dns failures. Oops.
164 - Prevent servers from publishing an internal IP as their address.
165 - Address a strcat vulnerability in circuit.c
168 Changes in version 0.0.2pre25 - 2004-03-04
170 - Put the OR's IP in its router descriptor, not its fqdn. That way
171 we'll stop being stalled by gethostbyname for nodes with flaky dns,
174 - If the user typed in an address that didn't resolve, the server
178 Changes in version 0.0.2pre24 - 2004-03-03
180 - Fix an assertion failure in dns.c, where we were trying to dequeue
181 a pending dns resolve even if it wasn't pending
182 - Fix a spurious socks5 warning about still trying to write after the
183 connection is finished.
184 - Hold certain marked_for_close connections open until they're finished
185 flushing, rather than losing bytes by closing them too early.
186 - Correctly report the reason for ending a stream
187 - Remove some duplicate calls to connection_mark_for_close
188 - Put switch_id and start_daemon earlier in the boot sequence, so it
189 will actually try to chdir() to options.DataDirectory
190 - Make 'make test' exit(1) if a test fails; fix some unit tests
191 - Make tor fail when you use a config option it doesn't know about,
192 rather than warn and continue.
193 - Make --version work
194 - Bugfixes on the rpm spec file and tor.sh, so it's more up to date
197 Changes in version 0.0.2pre23 - 2004-02-29
199 - Print a statement when the first circ is finished, so the user
201 - If a relay cell is unrecognized at the end of the circuit,
202 send back a destroy. (So attacks to mutate cells are more
204 - New config option 'excludenodes' to avoid certain nodes for circuits.
205 - When it daemonizes, it chdir's to the DataDirectory rather than "/",
206 so you can collect coredumps there.
208 - Fix a bug in tls flushing where sometimes data got wedged and
209 didn't flush until more data got sent. Hopefully this bug was
210 a big factor in the random delays we were seeing.
211 - Make 'connected' cells include the resolved IP, so the client
212 dns cache actually gets populated.
213 - Disallow changing from ORPort=0 to ORPort>0 on hup.
214 - When we time-out on a stream and detach from the circuit, send an
215 end cell down it first.
216 - Only warn about an unknown router (in exitnodes, entrynodes,
217 excludenodes) after we've fetched a directory.
220 Changes in version 0.0.2pre22 - 2004-02-26
222 - Servers publish less revealing uname information in descriptors.
223 - More memory tracking and assertions, to crash more usefully when
225 - If the default torrc isn't there, just use some default defaults.
226 Plus provide an internal dirservers file if they don't have one.
227 - When the user tries to use Tor as an http proxy, give them an http
228 501 failure explaining that we're a socks proxy.
229 - Dump a new router.desc on hup, to help confused people who change
230 their exit policies and then wonder why router.desc doesn't reflect
232 - Clean up the generic tor.sh init script that we ship with.
234 - If the exit stream is pending on the resolve, and a destroy arrives,
235 then the stream wasn't getting removed from the pending list. I
236 think this was the one causing recent server crashes.
237 - Use a more robust poll on OSX 10.3, since their poll is flaky.
238 - When it couldn't resolve any dirservers, it was useless from then on.
239 Now it reloads the RouterFile (or default dirservers) if it has no
241 - Move the 'tor' binary back to /usr/local/bin/ -- it turns out
242 many users don't even *have* a /usr/local/sbin/.
245 Changes in version 0.0.2pre21 - 2004-02-18
247 - There's a ChangeLog file that actually reflects the changelog.
248 - There's a 'torify' wrapper script, with an accompanying
249 tor-tsocks.conf, that simplifies the process of using tsocks for
250 tor. It even has a man page.
251 - The tor binary gets installed to sbin rather than bin now.
252 - Retry streams where the connected cell hasn't arrived in 15 seconds
253 - Clean up exit policy handling -- get the default out of the torrc,
254 so we can update it without forcing each server operator to fix
256 - Allow imaps and pop3s in default exit policy
258 - Prevent picking middleman nodes as the last node in the circuit
261 Changes in version 0.0.2pre20 - 2004-01-30
263 - We now have a deb package, and it's in debian unstable. Go to
265 - I've split the TotalBandwidth option into BandwidthRate (how many
266 bytes per second you want to allow, long-term) and
267 BandwidthBurst (how many bytes you will allow at once before the cap
268 kicks in). This better token bucket approach lets you, say, set
269 BandwidthRate to 10KB/s and BandwidthBurst to 10MB, allowing good
270 performance while not exceeding your monthly bandwidth quota.
271 - Push out a tls record's worth of data once you've got it, rather
272 than waiting until you've read everything waiting to be read. This
273 may improve performance by pipelining better. We'll see.
274 - Add an AP_CONN_STATE_CONNECTING state, to allow streams to detach
275 from failed circuits (if they haven't been connected yet) and attach
277 - Expire old streams that haven't managed to connect. Some day we'll
278 have them reattach to new circuits instead.
281 - Fix several memory leaks that were causing servers to become bloated
283 - Fix a few very rare assert triggers. A few more remain.
284 - Setuid to User _before_ complaining about running as root.
287 Changes in version 0.0.2pre19 - 2004-01-07
289 - Fix deadlock condition in dns farm. We were telling a child to die by
290 closing the parent's file descriptor to him. But newer children were
291 inheriting the open file descriptor from the parent, and since they
292 weren't closing it, the socket never closed, so the child never read
293 eof, so he never knew to exit. Similarly, dns workers were holding
294 open other sockets, leading to all sorts of chaos.
295 - New cleaner daemon() code for forking and backgrounding.
296 - If you log to a file, it now prints an entry at the top of the
297 logfile so you know it's working.
298 - The onionskin challenge length was 30 bytes longer than necessary.
299 - Started to patch up the spec so it's not quite so out of date.
302 Changes in version 0.0.2pre18 - 2004-01-02
304 - Fix endian issues with the 'integrity' field in the relay header.
305 - Fix a potential bug where connections in state
306 AP_CONN_STATE_CIRCUIT_WAIT might unexpectedly ask to write.
309 Changes in version 0.0.2pre17 - 2003-12-30
311 - Made --debuglogfile (or any second log file, actually) work.
312 - Resolved an edge case in get_unique_circ_id_by_conn where a smart
313 adversary could force us into an infinite loop.
316 - Each onionskin handshake now includes a hash of the computed key,
317 to prove the server's identity and help perfect forward secrecy.
318 - Changed cell size from 256 to 512 bytes (working toward compatibility
320 - Changed cell length to 2 bytes, and moved it to the relay header.
321 - Implemented end-to-end integrity checking for the payloads of
323 - Separated streamid from 'recognized' (otherwise circuits will get
324 messed up when we try to have streams exit from the middle). We
325 use the integrity-checking to confirm that a cell is addressed to
327 - Randomize the initial circid and streamid values, so an adversary who
328 breaks into a node can't learn how many circuits or streams have
332 Changes in version 0.0.2pre16 - 2003-12-14
334 - Fixed a bug that made HUP trigger an assert
335 - Fixed a bug where a circuit that immediately failed wasn't being
336 counted as a failed circuit in counting retries.
339 - Now we close the circuit when we get a truncated cell: otherwise we're
340 open to an anonymity attack where a bad node in the path truncates
341 the circuit and then we open streams at him.
342 - Add port ranges to exit policies
343 - Add a conservative default exit policy
344 - Warn if you're running tor as root
345 - on HUP, retry OR connections and close/rebind listeners
346 - options.EntryNodes: try these nodes first when picking the first node
347 - options.ExitNodes: if your best choices happen to include any of
348 your preferred exit nodes, you choose among just those preferred
350 - options.ExcludedNodes: nodes that are never picked in path building
353 Changes in version 0.0.2pre15 - 2003-12-03
354 o Robustness and bugfixes:
355 - Sometimes clients would cache incorrect DNS resolves, which would
356 really screw things up.
357 - An OP that goes offline would slowly leak all its sockets and stop
359 - A wide variety of bugfixes in exit node selection, exit policy
360 handling, and processing pending streams when a new circuit is
362 - Pick nodes for a path only from those the directory says are up
363 - Choose randomly from all running dirservers, not always the first one
364 - Increase allowed http header size for directory fetch.
365 - Stop writing to stderr (if we're daemonized it will be closed).
366 - Enable -g always, so cores will be more useful to me.
367 - Switch "-lcrypto -lssl" to "-lssl -lcrypto" for broken distributions.
370 - Wrote a man page. It lists commonly used options.
373 - Change default loglevel to warn.
374 - Make PidFile default to null rather than littering in your CWD.
375 - OnionRouter config option is now obsolete. Instead it just checks
377 - Moved to a single unified torrc file for both clients and servers.
380 Changes in version 0.0.2pre14 - 2003-11-29
381 o Robustness and bugfixes:
382 - Force the admin to make the DataDirectory himself
383 - to get ownership/permissions right
384 - so clients no longer make a DataDirectory and then never use it
385 - fix bug where a client who was offline for 45 minutes would never
386 pull down a directory again
387 - fix (or at least hide really well) the dns assert bug that was
388 causing server crashes
389 - warnings and improved robustness wrt clockskew for certs
390 - use the native daemon(3) to daemonize, when available
391 - exit if bind() fails
392 - exit if neither socksport nor orport is defined
393 - include our own tor_timegm (Win32 doesn't have its own)
394 - bugfix for win32 with lots of connections
395 - fix minor bias in PRNG
396 - make dirserver more robust to corrupt cached directory
399 - Wrote the design document (woo)
401 o Circuit building and exit policies:
402 - Circuits no longer try to use nodes that the directory has told them
404 - Exit policies now support bitmasks (18.0.0.0/255.0.0.0) and
405 bitcounts (18.0.0.0/8).
406 - Make AP connections standby for a circuit if no suitable circuit
407 exists, rather than failing
408 - Circuits choose exit node based on addr/port, exit policies, and
409 which AP connections are standing by
410 - Bump min pathlen from 2 to 3
411 - Relay end cells have a payload to describe why the stream ended.
412 - If the stream failed because of exit policy, try again with a new
414 - Clients have a dns cache to remember resolved addresses.
415 - Notice more quickly when we have no working circuits
418 - APPort is now called SocksPort
419 - SocksBindAddress, ORBindAddress, DirBindAddress let you configure
421 - RecommendedVersions is now a config variable rather than
422 hardcoded (for dirservers)
423 - Reloads config on HUP
424 - Usage info on -h or --help
425 - If you set User and Group config vars, it'll setu/gid to them.