[metrics-team] Tor Weather: data sources

[nusenu]

Damian Johnson:
> Thanks nusenu. I'm still unsure Onionoo makes sense but happy to leave
> it to whoever makes the service.
> 
>> - how long is the relay down? (last_seen timestamp)
> 
> Unless I'm missing something if Onionoo derives this from 'is relay x
> in the hourly consensus' then this is no different than 'download the
> hourly consensus and if the relay's gone then note it'. 

note that the required information is not binary ("running vs. gone")

but running for _how long_ because (old) Tor Weather supported

"only tell me if my relay is gone for ___ hours/days/weeks"
which is not possible without storing multiple consensuses or using 
onionoo's last_seen timestamp. 
Also with onionoo you will not be able to implement anything >7 days
of downtime since onionoo only shows relays seen at some point within the
last 7 days.

yes you can do that by implementing a last_seen timestamp using stem
but why reimplement what onionoo already does for you?

> You could get
> a more precise measure by active probing (as already discussed) but
> otherwise all this field provides you is information Weather already
> knows.

I don't think we are aiming for any precise (< 1hour granularity) measurement here.

 
> Gotcha. I'm unsure why Weather would require this but you're right
> that if it is this could be a reason to use Onionoo.

because relay operators will be able to subscribe to their family
without the need to subscribe to every single relay or adjust their
subscription if they add more relays to their family


-- 
https://twitter.com/nusenu_
https://mastodon.social/@nusenu

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 833 bytes
Desc: OpenPGP digital signature
URL: <http://lists.torproject.org/pipermail/metrics-team/attachments/20180720/07c6d633/attachment-0001.sig>