Tor 0.2.2.7-alpha released

alpha fixes a huge client-side performance bug, as well
as laying the groundwork for further relay-side performance fixes. It
also starts cleaning up client behavior with respect to the EntryNodes,
ExitNodes, and StrictNodes config options.

This release also rotates two directory authority keys, due to a security
breach of some of the Torproject servers:
http://archives.seul.org/or/talk/Jan-2010/msg00161.html

Everybody should upgrade:
https://www.torproject.org/download.html.en

Changes in version 0.2.2.7-alpha - 2010-01-19
o Directory authority changes:
- Rotate keys (both v3 identity and relay identity) for moria1
and gabelmoo.

o Major features (performance):
- We were selecting our guards uniformly at random, and then weighting
which of our guards we'd use uniformly at random. This imbalance
meant that Tor clients were severely limited on throughput (and
probably latency too) by the first hop in their circuit. Now we
select guards weighted by currently advertised bandwidth. We also
automatically discard guards picked using the old algorithm. Fixes
bug 1217; bugfix on 0.2.1.3-alpha. Found by Mike Perry.
- When choosing which cells to relay first, relays can now favor
circuits that have been quiet recently, to provide lower latency
for low-volume circuits. By default, relays enable or disable this
feature based on a setting in the consensus. You can override
this default by using the new "CircuitPriorityHalflife" config
option. Design and code by Ian Goldberg, Can Tang, and Chris
Alexander.
- Add separate per-conn write limiting to go with the per-conn read
limiting. We added a global write limit in Tor 0.1.2.5-alpha,
but never per-conn write limits.
- New consensus params "bwconnrate" and "bwconnburst" to let us
rate-limit client connections as they enter the network. It's
controlled in the consensus so we can turn it on and off for
experiments. It's starting out off. Based on proposal 163.

o Major features (relay selection options):
- Switch to a StrictNodes config option, rather than the previous
"StrictEntryNodes" / "StrictExitNodes" separation that was missing a
"StrictExcludeNodes" option.
- If EntryNodes, ExitNodes, ExcludeNodes, or ExcludeExitNodes
change during a config reload, mark and discard all our origin
circuits. This fix should address edge cases where we change the
config options and but then choose a circuit that we created before
the change.
- If EntryNodes or ExitNodes are set, be more willing to use an
unsuitable (e.g. slow or unstable) circuit. The user asked for it,
they get it.
- Make EntryNodes config option much more aggressive even when
StrictNodes is not set. Before it would prepend your requested
entrynodes to your list of guard nodes, but feel free to use others
after that. Now it chooses only from your EntryNodes if any of
those are available, and only falls back to others if a) they're
all down and b) StrictNodes is not set.
- Now we refresh your entry guards from EntryNodes at each consensus
fetch -- rather than just at startup and then they slowly rot as
the network changes.

o Major bugfixes:
- Stop bridge directory authorities from answering dbg-stability.txt
directory queries, which would let people fetch a list of all
bridge identities they track. Bugfix on 0.2.1.6-alpha.

o Minor features:
- Log a notice when we get a new control connection. Now it's easier
for security-conscious users to recognize when a local application
is knocking on their controller door. Suggested by bug 1196.
- New config option "CircuitStreamTimeout" to override our internal
timeout schedule for how many seconds until we detach a stream from
a circuit and try a new circuit. If your network is particularly
slow, you might want to set this to a number like 60.
- New controller command "getinfo config-text". It returns the
contents that Tor would write if you send it a SAVECONF command,
so the controller can write the file to disk itself.
- New options for SafeLogging to allow scrubbing only log messages
generated while acting as a relay.
- Ship the bridges spec file in the tarball too.
- Avoid a mad rush at the beginning of each month when each client
rotates half of its guards. Instead we spread the rotation out
throughout the month, but we still avoid leaving a precise timestamp
in the state file about when we first picked the guard. Improves
over the behavior introduced in 0.1.2.17.

o Minor bugfixes (compiling):
- Fix compilation on OS X 10.3, which has a stub mlockall() but
hides it. Bugfix on 0.2.2.6-alpha.
- Fix compilation on Solaris by removing support for the
DisableAllSwap config option. Solaris doesn't have an rlimit for
mlockall, so we cannot use it safely. Fixes bug 1198; bugfix on
0.2.2.6-alpha.

o Minor bugfixes (crashes):
- Do not segfault when writing buffer stats when we haven't observed
a single circuit to report about. Found by Fabian Lanze. Bugfix on
0.2.2.1-alpha.
- If we're in the pathological case where there's no exit bandwidth
but there is non-exit bandwidth, or no guard bandwidth but there
is non-guard bandwidth, don't crash during path selection. Bugfix
on 0.2.0.3-alpha.
- Fix an impossible-to-actually-trigger buffer overflow in relay
descriptor generation. Bugfix on 0.1.0.15.

o Minor bugfixes (privacy):
- Fix an instance where a Tor directory mirror might accidentally
log the IP address of a misbehaving Tor client. Bugfix on
0.1.0.1-rc.
- Don't list Windows capabilities in relay descriptors. We never made
use of them, and maybe it's a bad idea to publish them. Bugfix
on 0.1.1.8-alpha.

o Minor bugfixes (other):
- Resolve an edge case in path weighting that could make us misweight
our relay selection. Fixes bug 1203; bugfix on 0.0.8rc1.
- Fix statistics on client numbers by country as seen by bridges that
were broken in 0.2.2.1-alpha. Also switch to reporting full 24-hour
intervals instead of variable 12-to-48-hour intervals.
- After we free an internal connection structure, overwrite it
with a different memory value than we use for overwriting a freed
internal circuit structure. Should help with debugging. Suggested
by bug 1055.
- Update our OpenSSL 0.9.8l fix so that it works with OpenSSL 0.9.8m
too.

o Removed features:
- Remove the HSAuthorityRecordStats option that version 0 hidden
service authorities could have used to track statistics of overall
hidden service usage.