Hi NoisyCoil,
Sorry for the delay! As you'll see below, there was no quick and easy
answer to your questions.
NoisyCoil via Tails-dev (2025-10-28):
> A lot of time has passed since I last gave updates on the mailing list
> about the arm64 port of Tails. I've been quite busy at work and I still
> am (in fact, I am busier now than ever), but I've procrastinated writing
> this email for 6+ months now, so I thought I'd better start writing it
> now and finish whenever I finish :-) So here you go.
Thank you for the detailed update! Impressive work!
(And with my DD hat on: welcome to Debian :)
> I have one question for the Tails developers/infra maintainers.
> Is there any chance Tails could provide time-based and tagged
> snapshots for arm64 packages in addition to amd64? Would it be too
> expensive? That would greatly simplify the work of contributors (for
> instance, there was a request to the mailing list just a few days
> ago), as at this time a dirty DNS hack is required to make the build
> system install arm64 packages from Debian instead of from the Tails
> snapshot servers.
Your last status updates and in particular the specific request I've
quoted above raised a bunch of questions for us, because we were torn:
- On the one hand, you've demonstrated such a level of commitment to
the "Tails on arm64 + Apple Silicon" project that I feel you
deserve some support.
- But support *towards what*? Now that you have a working
proof-of-concept, which is sufficient to demonstrate feasibility to
some degree, we need clarity about what needs to happen for this
port to become an official *product* — and whether this is even
a realistic goal, that's worth investing Tor resources into (and
potentially seeking extra resources dedicated to this goal).
So, back in November, the Tails Team and sajolida have met and we've
discussed the ARM64 (and more specifically Apple Silicon) situation.
I'm hereby reporting back from these discussions and bringing
a proposal to the table.
I've saved a reorganized version of our notes on a new blueprint:
https://gitlab.tails.boum.org/tails/blueprints/-/wikis/ARM_platforms/ARM64
There are 2 mains parts to it:
What's the best UX we can reasonably hope for?
==============================================
In that section we tried to understand what a Tails product that
supports Apple Silicon machines would look like, in the best case
scenario, for our users.
This is useful to gauge whether it's even worth bothering: e.g. if it
turns out the best we can hope for is not usable enough for our target
users, or does not support the actual computers we're trying to bring
Tails to, then maybe this project is not worth our efforts (in which
case it's not necessarily the end of the road, e.g. it could
potentially live as a Tails derivative).
We did not reach conclusions but we gathered pretty significant
data already.
Can you please review our findings and fix them if needed? I'd like
this section to reflect fairly what it is about! (I'm happy to request
write access for you on the blueprints wiki, if you want to edit the
page directly.)
Proposal towards experimenting with Tails *alpha* releases for Apple Silicon
============================================================================
In order to gauge interest by users, and to get a better understanding
of the associated costs on our side, we had the idea to release,
during 12 months, e.g. from late 2026 to late 2027, *alpha* releases
for Apple Silicon, in a way that's as cheap as possible for the Tails
Team. It would rely on you doing most of the work, both in terms of
getting to the point when we can start this experiment, and in terms
of keeping the experiment alive.
There's quite some work that needs to happen until we can start but if
we can reach an agreement, we're prepared to commit to our part of it.
Please read this proposal and think about whether this makes sense to
you, and whether you would consider committing to it. Then I suggest
we have a video call to discuss your feedback and the next steps.
APT snapshots
=============
Finally, regarding your specific request about APT snapshots:
- I'm assuming the hardware & bandwidth cost won't be a blocker.
I did not check how much human effort all this requires (I'm not
too worried about time-based snapshots; but tagged snapshots are
a bit more complicated to integrate.) I can check all of that
if/once we're closer to deciding to do this.
- We won't go ahead with this change *in isolation*. But if we go
ahead with our proposal, it's indeed 1 of the blockers that we'll
need to address!
Cheers,
--
intrigeri
