Logforme m7527@abc.se wrote:
On 2021-05-26 08:18:32, "Scott Bennett" bennett@sdf.org wrote:
I interpret that as meaning that one or more criteria being used by one or more authorities has changed,
What I have noticed on my relay is that the "Consensus Weight" is fluctuating. CW is too complicated for my tiny brain but I believe the measurements from the Bandwidth Authorities is involved. The BWAuths are spread around the world and depending on current internet conditions they get different speed values to your relay. But can it cause massive swings in CW?
Yes. My relay is on a residential service connection and its "bandwidth" rating from the Authority relays typically oscillates between ~30 and ~120, so proportionally the swings are often fairly wild.
Maybe the BWAuths have changed their measurement technique during the last couple of months?
Well, I first noticed it late last year, IIRC. The measurement technique will, of course, often give deceptive results. For example, if the connection supports ~350 KB/s and the relay has little traffic at the time measurement begins, the result should be fairly close to the true value. OTOH, if the relay is handling 200 KB/s of traffic for other circuits at the time measurement begins, then the result should be at most only ~150 KB/s, which is far from the true value.
A further question I would like to raise is why do the Authority relays use different criteria from one to another for the automatic assignment of flags? Should they not be consistent in using the same rules?
I agree that it is confusing that 2 auths don't assign the HSDir flag according to the spec. I have no explanation apart from that AFAIK moria1 is run by Roger Dingledine and I guess he runs a lot of beta and test stuff. moria1 publishes 2 HSDir "Flag Threshold" values (hsdir-wfu and
Yeah, I saw that, but don't know quite what to make of it.
hsdir-tk) that no other auth publishes which leads me to believe moria1 runs another version of the auth software that handles the HSDir flag differently. That don't explain bastet though.
And it only accounts for two Authority relays, whereas you said five are refusing to assign HSDir to my relay, which, as you pointed out, may depend upon network conditions between those Authority relays and my relay at the time and have nothing at all to do with my relay or how much traffic my relay could handle or might actually be handling at the time.
It's fun to speculate :)
I would rather not be kept in the dark. It should not be like trying to get information on what the criminals who rule over us are up to. The problems outlined above would be mitigated somewhat if the measurements were filtered somehow, which could be as simple a filter as a boxcar moving average. Yes, I know that for many purposes a rectangular window gives lousy results, but for the purpose of understanding relays' capacities over time as having values that usually change slowly if at all a boxcar moving average should be plenty good enough. An exponential moving average would probably also be fine. The point of using a filter for the measurements would be to minimize the temporary interference of transient network conditions affecting the measurement process and corrupting some of the results at some times but not at others. Of course, measurements outside some number of standard deviations from the mean for a relay could be discarded, as well. At present, it is difficult to separate the deficiencies of the measurement method from the network realities in trying to interpret the measurements.
Scott Bennett, Comm. ASMELG, CFIAG ********************************************************************** * Internet: bennett at sdf.org *xor* bennett at freeshell.org * *--------------------------------------------------------------------* * "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 * **********************************************************************