1 Filename: 148-uniform-client-end-reason.txt
2 Title: Stream end reasons from the client side should be uniform
5 Author: Roger Dingledine
8 Implemented-In: 0.2.1.9-alpha
12 When a stream closes before it's finished, the end relay cell that's
13 sent includes an "end stream reason" to tell the other end why it
14 closed. It's useful for the exit relay to send a reason to the client,
15 so the client can choose a different circuit, inform the user, etc. But
16 there's no reason to include it from the client to the exit relay,
17 and in some cases it can even harm anonymity.
19 We should pick a single reason for the client-to-exit-relay direction
20 and always just send that.
24 Back when I first deployed the Tor network, it was useful to have
25 the Tor relays learn why a stream closed, so I could debug both ends
26 of the stream at once. Now that streams have worked for many years,
27 there's no need to continue telling the exit relay whether the client
28 gave up on a stream because of "timeout" or "misc" or what.
30 Then in Tor 0.2.0.28-rc, I fixed this bug:
31 - Fix a bug where, when we were choosing the 'end stream reason' to
32 put in our relay end cell that we send to the exit relay, Tor
33 clients on Windows were sometimes sending the wrong 'reason'. The
34 anonymity problem is that exit relays may be able to guess whether
35 the client is running Windows, thus helping partition the anonymity
36 set. Down the road we should stop sending reasons to exit relays,
37 or otherwise prevent future versions of this bug.
39 It turned out that non-Windows clients were choosing their reason
40 correctly, whereas Windows clients were potentially looking at errno
41 wrong and so always choosing 'misc'.
43 I fixed that particular bug, but I think we should prevent future
44 versions of the bug too.
46 (We already fixed it so *circuit* end reasons don't get sent from
47 the client to the exit relay. But we appear to be have skipped over
48 stream end reasons thus far.)
52 One option would be to no longer include any 'reason' field in end
53 relay cells. But that would introduce a partitioning attack ("users
54 running the old version" vs "users running the new version").
56 Instead I suggest that clients all switch to sending the "misc" reason,
57 like most of the Windows clients currently do and like the non-Windows
58 clients already do sometimes.