[Author Prev][Author Next][Thread Prev][Thread Next][Author Index][Thread Index]
[tor-bugs] Re: #1690 [Tor - Relay]: Consensus Bandwidth Lacks Indication of Type
#1690: Consensus Bandwidth Lacks Indication of Type
---------------------------------+------------------------------------------
Reporter: atagar | Owner:
Type: enhancement | Status: new
Priority: minor | Milestone:
Component: Tor - Relay | Version:
Keywords: consensus bandwidth | Parent:
---------------------------------+------------------------------------------
Comment(by atagar):
Here's the irc discussion of this topic (heavily trimmed):
07:49 < atagar> In the network status' bandwidth ('w') entry the third
value is the observed bandwidth, but iirc from a past discussion it's the
client side measured bandwidth for older versions of tor, right?
07:50 < karsten> there are 3 cases:
07:51 < karsten> https://trac.torproject.org/projects/tor/ticket/1566
...
08:11 < atagar> karsten: How can you tell on the client side which
measurement it is? For instance if an entry is "w Bandwidth=65700" how do
I tell if this is observed, measured, or weighted measured?
08:12 < karsten> observed or measured. weighting happens on client side.
08:12 < karsten> i count the votes that have Measured lines in them. if
there are 3+ such votes for a consensus, values are measured. otherwise
observed.
08:12 < karsten> fun, isn't it? :)
...
08:35 < Sebastian> atagar: can't you just use the value from the consensus
always?
08:35 < Sebastian> What's wrong with that?
08:35 < atagar> it'll be mislabeled (I'm not sure if it's observed,
measured, or weighted measured)
08:36 < Sebastian> you could call it bandwidth ;)
08:36 < Sebastian> if you don't like it, I wonder why you don't file a bug
asking that there be another line added to the consensus to let the client
know; and rather want to fetch all votes
...
08:37 < atagar> I don't want to fetch all the votes (that would be
hideous) and putting extra information in the network status just for
display purposes would also be quite foolish
08:38 < atagar> though it is a pitty the bandwidth entry doesn't have a
single character flag to indicate what it represents...
08:39 < Sebastian> I think it would be easy to add a flag at the top of
the consensus, and it would be useful for stats, too
08:40 < atagar> karsten: thoughts on the flag idea?
08:40 < karsten> actually, i wondered why there is no flag.
08:40 < karsten> it takes a new consensus version, though.
08:41 < Sebastian> Right
08:41 < Sebastian> but consensus versions are easy
08:41 < atagar> change log for dir-v4 - we added a single character!
08:41 < Sebastian> erm... no
08:42 < karsten> v3 dir-spec, but v9/10/? consensus version
08:43 < karsten> that requires all dir auths to upgrade
08:43 < Sebastian> it's an easy thing to merge into the next consensus
version update
08:43 < karsten> that's true
08:44 < armadev> atagar: to make things even more fun, the weights in the
consensus are weights, not measured bandwidth. the bwauthorities come up
with numbers that overshoot, in the case of a fast relay that's not
getting enough attention. so we'll publish numbers like 60MB/s when we
didn't actually measure that you could do 60MB/s. what we measured is that
you can do what you're currently advertising way better than the other
people who advertise that same amount.
08:44 < Sebastian> (I expect there's be some more events where we need to
upgrade all authorities at once before 0.2.2.x becomes stable, anyways.
--
Ticket URL: <https://trac.torproject.org/projects/tor/ticket/1690#comment:1>
Tor Bug Tracker & Wiki <https://trac.torproject.org/>
The Tor Project: anonymity online