[Author Prev][Author Next][Thread Prev][Thread Next][Author Index][Thread Index]
Re: [tor-bugs] #11211 [Core Tor/Tor]: Multiple ServerTransportListenAddr entries should be allowed per transport.
#11211: Multiple ServerTransportListenAddr entries should be allowed per transport.
-------------------------------------------------+-------------------------
Reporter: yawning | Owner: kaie
Type: enhancement | Status:
| needs_revision
Priority: Medium | Milestone: Tor:
| 0.3.0.x-final
Component: Core Tor/Tor | Version: Tor:
| unspecified
Severity: Normal | Resolution:
Keywords: tor-bridge, pt, needs-proposal, | Actual Points:
tor-pt, bridgedb-parsers, 028-triage, ipv6, |
tor-03-unspecified-201612 |
Parent ID: #10629 | Points:
Reviewer: | Sponsor: T/U
-------------------------------------------------+-------------------------
Comment (by yawning):
Replying to [comment:22 dcf]:
> This is a good idea, but I think it's more complicated than just giving
a list to ServerTransportListenAddr. You would need to also make
ServerTransportOptions be similarly split, which would probably require
new syntax in torrc. It would also require a change to pt-spec, because
there would need to be a rule or something that states which options
pertain to which listening address when there are multiple ones.
obfs4proxy would require fairly substantial changes to handle this
correctly, as would core tor and probably the bridge distribution stuff.
So while it's a good idea, getting it to bind to addresses is just the
start, and even that needs to be preceded by a spec change so that code
that supports this use case will work, and that code that doesn't will
error out per the existing spec.
> I've been frustrated by this in the past, too. For example, in
comment:10:ticket:20348, I wanted to run three obfs4 bridges with slightly
different configuration on the same IP address, and there's just no way to
do it other than to run three instances of tor. It was probably a mistake
for torrc to use the transport name as a key that links
ServerTransportListenAddr and ServerTransportOptions, because that makes a
built-in assumption that there's only one thing identified by that
transport.
Lots of aspects of the design aren't great.
> Incidentally, it might be the the case that using only the IPv6 syntax
already does what you want. On some systems `[::]` will listen on both
IPv6 and IPv4, so you don't need to separately list 0.0.0.0.
That's a general Linux-ism. Only one address will be distributed via
BridgeDB though (fairly sure it's the IPv4 one).
--
Ticket URL: <https://trac.torproject.org/projects/tor/ticket/11211#comment:23>
Tor Bug Tracker & Wiki <https://trac.torproject.org/>
The Tor Project: anonymity online
_______________________________________________
tor-bugs mailing list
tor-bugs@xxxxxxxxxxxxxxxxxxxx
https://lists.torproject.org/cgi-bin/mailman/listinfo/tor-bugs