[LEDE-DEV] anonymous stats / was: Release Candidate Test Plan - first draft

Etienne Champetier champetier.etienne at gmail.com
Tue Feb 7 08:39:53 PST 2017


2017-02-07 7:40 GMT-08:00 Bastian Bittorf <bb at npl.de>:
> * Etienne Champetier <champetier.etienne at gmail.com> [07.02.2017 16:27]:
>> not a fan of leaking revision number on public network by default (if
>> you are connected to public wifi or ...)
>
> This is a valid point somehow, but:
> because model and revision number can/must be encoded with e.g. base64
> it is at least not plaintext, but far from encrypted 8-)

maybe we can use ROT26 :D

>
> Is it really an issue sending the revision-number/model over wire?
> Is an opt-out (uci-var) thinkable?
>
>> If someone do something, can this be over https?
>
> it's about sending via DNS, so no http/https involved here.

DNS is not meant to transport arbitrary data, you have many
restriction that you will quicly hit (63 octets per label, 255 total)
The 2 advantages of DNS are pseudo anonymity (the LEDE team will not
have the IPs of the routers) and firewall "bypass"
The disavantage is that you can't know how many routers, if someone
tries to mess with you metric, ...

I would prefer to start a real telemetry service, sending lots of info
(a unique id/board/revision/allpackages versions/...) in an extensible
format (json?)
put that behind an nginx with agressive rate limiting, drop the IP
information, and maybe make this data public

Haven't looked how debian and others are doing telemetry

Regards
Etienne

>
> bye, bastian



More information about the Lede-dev mailing list