Path Selection

[Damian Johnson]

Hopefully I've got these right...

Are the choices of the ORs in the path all based on the advertised bandwidth
> (Is that the average bandwidth value or the observed)?

At the moment weighting is based on advertised bandwidth (with the exception
of the entry and exit selection which has special constraints, of course).
Mike's currently looking into using observed bandwidth instead.


Do the directory servers select which nodes can be Guards (Is there a
> criteria for this)?

Yes. The stable relays are favored to be guards (there might be some other
criteria too...).


Does the individual client select a few Guards from the valid list of Guards
> and then only use them as the entry nodes?

Yes.


An OR being marked as Stable is really just a matter of it staying up and
> connected to the tor network for long enough?

Mostly. There's other considerations, for instance if accounting limit is
enabled then you won't be eligible for the stable flag.


Cheers! -Damian

On Tue, Jan 19, 2010 at 4:20 AM, Jason Cooper <J.L.Cooper at lboro.ac.uk>wrote:

>  Hi,
>
>   I am trying to get my head round the basics of the path selection for
> circuits.  I am putting together a simple simulator for tor networks to help
> me understand it and see what effects changing OR values has on the circuits
> built.  I have read
> http://gitweb.torproject.org/tor/tor.git/blob/HEAD:/doc/spec/path-spec.txtbut I have a few questions.
>
>
>
>   Are the choices of the ORs in the path all based on the advertised
> bandwidth (Is that the average bandwidth value or the observed)?
>
>
>
>   Do the directory servers select which nodes can be Guards (Is there a
> criteria for this)?
>
>
>
>   Does the individual client select a few Guards from the valid list of
> Guards and then only use them as the entry nodes?
>
>
>
>   An OR being marked as Stable is really just a matter of it staying up and
> connected to the tor network for long enough?
>
>
>
> Regards,
>
>   Jason.
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.torproject.org/pipermail/tor-talk/attachments/20100119/357bcc48/attachment.htm>