Re: TorStatus version 4 has been released

[Roger Dingledine <arma@xxxxxxx>]

On Mon, Jan 21, 2008 at 03:59:38PM -0700, Kasimir Gabert wrote:
> Hello all [especially TorStatus mirror operators],
> 
> A new version of TorStatus (http://torstatus.kgprog.com) has been
> released.  More information can be found at
> http://project.torstatus.kgprog.com, and any comments are greatly
> welcome.

Hi Kasimir,

Here's a wishlist that's been slowly accumulating in my head. :)

1) It looks like it's still using the v2 networkstatus format. (I'm
guessing this because there I sometimes see the "partially listed"
icon by a few relays.) The v3 networkstatus format also works now:
if you run a recent 0.2.0.x version, you may find the cached-consensus
file to be easier to look at than looking through each cached-status
file and computing locally. See also
https://www.torproject.org/svn/trunk/doc/spec/dir-spec.txt

2) I love the "bandwidth bar" in the relay list. But it would be more
meaningful to me if it listed average sustained load, not just peak
load. You can compute this daily average by adding up all the numbers
in the write-history and read-history lines (either in the descriptor
or in the extrainfo document, depending on the relay's Tor version)
and dividing by 2*86400. (This penalizes relays that don't list a whole
day's worth of bandwidth histories, but I claim that's a feature.) You
could leave the peak bandwidth number in place, for people who want that,
and then make the bandwidth bar and the number two separate columns.

3) It looks like it's still using the DNSEL server that Joe added in?
Compare to the "tor checker" on xenobite (https://torcheck.xenobite.eu/)
or the one that tup wrote that Jake is running on
http://check.torproject.org/ that does active measurements to notice
Tor relays that exit on a different IP address than they advertise.

4) The network_detail.php page includes non-Running servers in its
graphs. I think this is skewing the data, and in any case it's hard
to adjust for that in my brain. We've been thinking about removing
non-Running servers from the networkstatus consensus. But even if we
don't get around to doing that for a while, would it make sense to
discard them from the graphs?

Thanks!
--Roger