Hi Sebastian,
Thanks for responding.
On Sat, 21 Aug 2010 15:17:19 +0200 Sebastian Hahn <mail@xxxxxxxxxxxxxxxxx
>
wrote:
On Aug 21, 2010, at 1:49 PM, Scott Bennett wrote:
Yesterday (Friday) the number of relays given HSDir flags by the
directory authorities was reduced by a bit more than 50% from only a
day earlier. My node, MYCROFTsOtherChild, was and still is among
the
bereft. This same situation happened a couple of weeks ago, also
affecting my node. That time the flag reappeared several hours
after
I first noticed what had happened. This time the flag is still
missing.
It is not obvious to me why the HSDir flag has been stripped from
so many relays, especially relays that have, like mine, been up and
running for weeks since the last restart. Has a bug been
triggered in
the authority code?
Apparently enough authorities have upgraded to 0.2.2.15-alpha. See
the changelog for details.
Yes, I see that entry now in the Changelog, although I don't see
the
reasoning behind it. HSDir requests differ in just about every
imaginable
way from ordinary directory requests. The protocol is different, the
HSDir entry format is different, the request frequency is lower by
orders
of magnitude, the requests are--or at least as I recall--intended to
be
handled over internal circuits rather than via a DirPort, and so
on. I
just don't see any reason to tie the HSDir service to the presence/
absence
of a DirPort.
Also, given the switch from decentralized traffic distribution to
centralized traffic distribution in recent tor versions, nodes
advertising
throughput capacities of, say, 49 KB/s or less (I don't really know
where
the curve becomes steeper) get little use already. Given that
DirPort is
ignored when the advertised throughput capacity is less than 50 KB/
s, tying
HSDir service to the presence/absence of an active DirPort means that
relatively unused relays cannot even help out with handling the
trivial
traffic volume of HSDir requests. Doesn't that seem rather
counterproductive?