Despite the fact that we tell Firefox to use SOCKS for FTP, it appears as if it is broken. This probably is a
firefox bug, but filing it anyway to remind myself to look deeper into it later. I should test this with an
ssh socks proxy and see what happens.
[Automatically added by flyspray2trac: Operating System: All]
To upload designs, you'll need to enable LFS and have an admin enable hashed storage. More information
Child items ...
Show closed items
Linked items 0
Link issues together to show that they're related.
Learn more.
This may actually work now in TorBrowser. Upping priority as a reminder to test.
Trac: Description: Despite the fact that we tell Firefox to use SOCKS for FTP, it appears as if it is broken. This probably is a
firefox bug, but filing it anyway to remind myself to look deeper into it later. I should test this with an
ssh socks proxy and see what happens.
[Automatically added by flyspray2trac: Operating System: All]
to
Despite the fact that we tell Firefox to use SOCKS for FTP, it appears as if it is broken. This probably is a
firefox bug, but filing it anyway to remind myself to look deeper into it later. I should test this with an
ssh socks proxy and see what happens.
[Automatically added by flyspray2trac: Operating System: All] Priority: minor to major Component: Torbutton to TorBrowserButton Keywords: N/Adeleted, N/Aadded Actualpoints: N/AtoN/A Parent: N/AtoN/A Milestone: N/AtoN/A Points: N/AtoN/A
wanoskarnet points out that the ftp protocol may be not as compatible with tor as we might like. With passive mode, you end up making another connection to the ftp server. What if that other connection comes from a different circuit than the first? A properly behaving ftp server should be suspicious.
I wonder how networks like aol (which keep rotating you around IP addresses too) deal with that issue. Probably by handling it specially on their nat.
I'm thinking that TrackHostExits might be a viable workaround here. It would seem that you typically don't need to ftp to that many different sites, so setting it for the individual sites would work - or a second tor instance could be used.
Sadly, it looks like TrackHostExits cannot apply to ports.. We probably need two ways to specify ports:
*:port - All requests to port should attempt to use the same exit IP.
?:port - All requests to pairs of (IP, port) should use the same exit IP.
Or, we could just implement Robert Hogan's version of Proposal 171 for auto-newnyming on different ports. Supposedly a patch exists for that somewhere right now (does it have a ticket?)
Robert's patch might not enforce the use of the same exit IP by default though.. He may have written it to only concern itself with using a new circuit on new ports.
Trac: Version: 1.2.4 toN/A Component: TorBrowserButton to Tor Check
Do you have an example FTP site that is broken? And are we correct in our guess that it is circuit-switching that is causing the problem? Does TrackHostExits fix it for you?
All we have are guesses at this point and it is hard to devote resources to fixes based on guesses. It often ends up wasting time.
Setting 'TrackHostExits ftp.mydomain.tld' in ~/.torrc fixed ftp for me. This is a great workaround.
PLEASE BE AWARE that ftp sends your password in plaintext and my understanding is that your exit node will gain full access to your ftp server unless you use one of the TLS extensions to ftp.
Of course, it would be great if tor did this more intelligently to correlate these transactions less. E.g., enabling a specific flag to temporarily track exit nodes for a host that is accessed on port 21 with a timeout.
Update, sorry, TrackHostExits does not work. When it does work, FTP works fine ... but after a while Tor 0.2.3.25 (git-3fed5eb096d2d187) ignores it for me and again opens data connections on separate circuits than the control connection. Using IP instead of hostname (since data connections are opened via ip from server), connecting only via IP, HUPping tor, does not resolve it. This is unfortunate. Restarting tor temporarily fixes it, but it soons breaks again.