[tor-talk] fetching all server descriptors

[tagnaq]

> It's correct that metrics-db does not fetch non-referenced descriptors.
> But it does parse the cached-descriptors[.new] files from gabelmoo, one of
> the directory authorities.  So, the archives at least contain the
> descriptors published to gabelmoo.  But there may be descriptors that are
> only published to one or more of the other directory authorities which
> would then not be contained in the metrics archive.  I can't say how many
> descriptors we're talking about.
> 
> See #2763 and #2282 for more discussion about this issue.  The current
> status of these tasks is that we need to extend Tor to make the rejected
> descriptors and those not contained in the consensus available via the
> directory protocol.  Once that's done, I can extend metrics-db to archive
> these descriptors, too.  I'd like to turn off rsyncing the
> cached-descriptors* files from gabelmoo sooner rather than later.
> 
>> > If my assumption is correct:
>> > Are there also archives that contain all descriptors? (referenced +
>> > unreferenced)
>> > Does the directory-archive script[3] archive/fetch all descriptors?
> The directory-archive script only downloads referenced descriptors, too.

Thanks for your explanation.