Tails

#1 Updated by intrigeri 2014-07-24 20:23:55

We have one. I don’t think its content and way of working are considered as public info (yet?), though.

#2 Updated by ioerror 2014-07-24 21:02:17

I started to write such a script - shall I abandon it or should we publish both? Or something else?

#3 Updated by Anonymous 2014-07-24 21:52:21

I think there might be a misunderstanding.

Today, we tried to check 29 mirrors for faulty ISO images which have been reported by users - by hand.

That is why i suggested to write a script, using only publicly available data (host dl.a.b.o => download iso, check size, checksum, gpg verify, dump all this to a text file).

So, i suppose that the existing script contains information which is not public, and thus might stay secret :) However, it’d be cool to be able to verify the mirrors independently in a case like the one which i just described. Or am i missing something here?

#5 Updated by intrigeri 2014-07-24 23:07:07

>

> [...]
> Fetching Tails GnuPG signature from 83.212.104.246:
> 83.212.104.246.iso.sig file size: 1099026432
> [...]
> 

This seems to be a very large detached signature. Bug?

#6 Updated by ioerror 2014-07-24 23:11:29

Whoops, I had a typo - I’ll update the patch.

#10 Updated by intrigeri 2014-07-25 10:00:14

> We already have this.

Some “we” has it, but as demonstrated on this ticket, there’s a need for other people to have it too. Especially when they’re taking care of something “we” have not been able to do in a timely manner in the last few days, causing much user confusion, and quite a lot of user support overhead on IRC. So, IMO, closing this ticket as Resolved just does not work.

> It’s called `check-mirrors` and is coded in Ruby.
> But it’s too ugly to be hosted on a public repo :)

I don’t think the ugliness of the code is a good enough reason, in itself, to go on hiding it.

The historical reason to keep some obscurity around this topic is that we would like to make it slightly harder for a malicious mirror operator to 1. know that we’re checking what they are serving; 2. know exactly how, when, and from what IPs we are doing it.

However, the setup used to run this script is now documented in a public Puppet module, so I don’t think these reasons really hold anymore. Also, it would be good to allow others to use (and possibly improve) the code, and demonstrated on this ticket. So, I’m now in favour of publishing our script.

With all this in mind, sajolida, what do you think?

#11 Updated by ioerror 2014-07-25 12:15:52

I would encourage merging my patch, keeping your script private and if people need to check the mirror - they won’t have to recreate the work I’ve already done.

#12 Updated by sajolida 2014-07-26 07:44:58

> Especially when they’re taking care
> of something “we” have not been able to do in a timely manner in the
> last few days, causing much user confusion, and quite a lot of user
> support overhead on IRC.

I obviously missed something here as I don’t know what you are referring
to. Would you mind being more explicit?

>> But it’s too ugly to be hosted on a public repo :)
>
> I don’t think the ugliness of the code is a good enough reason, in
> itself, to go on hiding it.

That was a joke.

> With all this in mind, sajolida, what do you think?

Fine with me. Still, we can’t just make that repo public because of two
things:

- A hidden place for our ISO image is documented there. It was thought
as a censorship circumvention tool. We never used it and the script is
broken on this. But still, we would totally loose the point of having
this trick available.

- The repo also has pool administration tools. Yeah, that was my idea
in the first place… But maybe we don’t want this to published in the
same place.

#13 Updated by intrigeri 2014-07-26 08:03:46

>> Especially when they’re taking care
>> of something “we” have not been able to do in a timely manner in the
>> last few days, causing much user confusion, and quite a lot of user
>> support overhead on IRC.

> I obviously missed something here as I don’t know what you are referring
> to. Would you mind being more explicit?

Sure :) There are many download errors reported recently.

Maybe it’s only caused by more people trying to download Tails, maybe some mirrors are still serving the “old” 1.1 ISO, maybe something else is wrong.

>> With all this in mind, sajolida, what do you think?

> Fine with me. Still, we can’t just make that repo public because of two
> things:

Good points!

> - A hidden place for our ISO image is documented there. It was thought
> as a censorship circumvention tool. We never used it and the script is
> broken on this. But still, we would totally loose the point of having
> this trick available.

Ah, right. Then, the script could be adapted to load private configuration from a file that’s not in the same Git repo. Would need another ticket.

> - The repo also has pool administration tools. Yeah, that was my idea
> in the first place… But maybe we don’t want this to published in the
> same place.

Indeed, this would need to be split out too.

I say no emergency at all, but it would be good to do this all at some point :)

#14 Updated by sajolida 2014-07-26 09:19:12

> Sure :) There are many download errors reported recently.
>
> Maybe it’s only caused by more people trying to download Tails, maybe some mirrors are still serving the “old” 1.1 ISO, maybe something else is wrong.

Yes, I’ve been reported through email quite a few download errors since
1.1 as well. But none of the reports from check-mirrors since 1.1
reported an error on my side. Note that I still receive only one and not
two, cf Bug #7485.

So maybe those were only due to the fact that, for the first time in
months, all our users had to download full ISO images. But still, that
doesn’t sound like a good explanation…

Were people able to actually spot faulty mirrors?

>> - A hidden place for our ISO image is documented there. It was thought
>> as a censorship circumvention tool. We never used it and the script is
>> broken on this. But still, we would totally loose the point of having
>> this trick available.
>
> Ah, right. Then, the script could be adapted to load private configuration from a file that’s not in the same Git repo. Would need another ticket.

Created Feature #7666.

>> - The repo also has pool administration tools. Yeah, that was my idea
>> in the first place… But maybe we don’t want this to published in the
>> same place.
>
> Indeed, this would need to be split out too.

Created Feature #7667.

Now closing that ticket.

Sorry, but I don’t know how to mark related tickets through email…

#15 Updated by intrigeri 2014-07-26 09:27:08

> Were people able to actually spot faulty mirrors?

Asking (Cc’d) the people who worked on it.

#16 Updated by Anonymous 2014-07-27 14:07:15

intrigeri wrote:
> > Were people able to actually spot faulty mirrors?
>
> Asking (Cc’d) the people who worked on it.

Just for the sake of adding this info to the ticket: we tested only 22 mirrors. None of them had a faulty image.