Re: 0.0.9.3 and hidden services

[Roger Dingledine <arma@xxxxxxx>]

On Thu, Jan 27, 2005 at 09:35:16AM +0100, Janos.Farkas-lists+priv-#jbaSOFUBavB*-or-talk@xxxxxxxxxxxxxxxxx wrote:
> Is 0.0.9.3 really making "hidden services less unbearable" for anyone
> else? :)

Heh. :)

I only backported a few of the hidden service fixes I made in the 0.1.0
cvs. The cvs code works much better with respect to hidden services
(for me at least), but cvs currently also has only half-implemented
libevent support, so it's unstable for now.

As for the performance problems lately, it looks like there are a lot
of variables going at once. One of them is that we've been having some
churn in the server lists (getting some new servers, losing some old
ones). Another is that it's the end of the month, so some of the nodes
mis-estimated how much bandwidth they'd need for the month and they're
hibernating now. A third is that I think we're seeing a lot more bytes
passed over the network now than, say, a week ago. And finally, as we
added more cable modem and dsl nodes, my bandwidth capacity detection
algorithm is working less well, meaning clients are sent to cable modem
nodes more often than they should for proper load balancing.

So, don't worry, we're still here, but for now Nick and I are working
on design and writing down some of our ideas, in preparation for the
upcoming PET submission deadline (www.petworkshop.org). In the meantime
we are piling up the bug reports at http://bugs.noreply.org/tor so they
don't slip through the cracks. We're betting that the Tor network won't
entirely collapse by the time we get back to coding and can start to
address these issues.

Thanks,
--Roger