[Author Prev][Author Next][Thread Prev][Thread Next][Author Index][Thread Index]

Re: Switching Tor relay speed

     On Sun, 22 Feb 2009 23:20:05 +0100 Lexi Pimenidis
<lexi@xxxxxxxxxxxxxxxxxxxxxxxxxxxx> wrote:
>On Sun, Feb 22, 2009 at 11:01:46PM CET, Scott Bennett wrote:
>>      No, no, NO.  That means three times as many descriptor uploads to the
>> authorities, which is the small part, and then potentially twice as many
>> fetches of your relay's descriptor by every relay and active client around the
>> planet during the next couple of hours.  That is exactly the wrong approach.
>> Please read my previous followup.
>I read it and I'm not convinced that the uploads of the status are so much

     The status file doesn't get sent anywhere.  We were discussing descriptor

>worse than having a multiplicity of clients "knowing" the wrong bandwidth.

     In general, it is a small problem compared to the others involved in
data rate reporting.  It is also one that is automatically corrected after
the final step of your three-step process occurs.  Only the one step is

>Just to make clear: I'm not proposing that all routers should to more
>updates, just in this single case it might be better. 

     First off, at most, two of the three steps will be picked up by any
instance of tor when they occur so close together in time, so one goes to
waste anyway.  Secondly, different instances of tor will get different single
descriptors of pairs of descriptors out of the three that your process will
upload, so there will be inconsistency of the information across the tor
network for one to three hours anyway, so there's no real improvement by
doing it three times instead of one.  The way the code is written, a three-
step process should do it by increments/decrements of >50% from the previous
descriptor anyway to avoid having the new descriptor ignored by the
authorities, IIRC.
>Also: if the authorities (or for the matter: the complete directory system)
>go(goes) down just because a few more updates come in, then the complete
>directory system should be revised ASAP because it will break when the next
>few nodes will join in.
     There were obviously many reasons for the succession of directory
protocol revisions, but performance and load on the tor network needed to
handle directory information appear to have been central reasons.  (See, for
example, the shuffling off of data not actually used for anything by tor
into extra-info documents.)  However, it looks like the problems with data
rate reporting will require yet another update to the protocol in order to
fix them. :-(

                                  Scott Bennett, Comm. ASMELG, CFIAG
* Internet:       bennett at cs.niu.edu                              *
* "A well regulated and disciplined militia, is at all times a good  *
* objection to the introduction of that bane of all free governments *
* -- a standing army."                                               *
*    -- Gov. John Hancock, New York Journal, 28 January 1790         *