Move the verification JavaScript from the verification extension to the page itself

Nothing in https://tails.boum.org/contribute/design/verification_extension/ actually prevents us from including the code to verify Tails ISO images directly on our website. The extension does not seem to provide any particular benefit (since it’s been rewritten at least) over using the forge library and the verification code directly in the browser. As we use native Web APIs from the browser, we have access to the File objects and could even create a drag & drop area for the files to check.

The extension does not protect from MITM (then everything would be MITM’ed) and it does not protect from phishing (somebody who creates a webpage called tails.boum.org using weird unicode can always have a lets encrypt certificate for their domain and
pretend that everything is working well).

I’ve re-read the abovementioned design document and I cannot see any benefit in maintaining an extension currently.

Here is a proposal of iterations to get this done:

Iteration #1

• have a security discussion
• if we agree on security: move the code to the website without
  changing the current setup (we would change some button texts but
  not fundamentally modify the layout and functioning of this page)
  This means keeping the same interactions:

# The user downloads the file
# The user clicks on “Verify Tails 3.13…”

- implementation-wise this could be done in a prototype at first
- review and modify carefully our Cross-origin policies, so that thread H (in the blueprint) is mitigated.
- Check if caching of JS files on the server poses a problem

• properly deprecate the extension and make people aware that it’s
  unsupported and it does not fire on the page (need to research
  mechanisms to do that)
• Update the design documentation

At this point we’ll already be able to see the benefits of the
replacement in terms of usage on the number of downloads. We’ll also be
in a better position to guess how much improvement we can hope from the
on-the-fly verification.

We could also

• have an unreleased proof-of-concept on the feasibility of on-the-fly
  verification. So we’re better armed to evaluate the cost of a second
  iteration.

Hourly budget estimation: 60 hrs in total for everyone involved

Iteration #2

• revamp visually + user testing
• implement on-the-fly verification
• if needed at this point: more code

Hourly budget estimation: 40 hrs in total for everyone involved

Related issues

• Related to #15995 (closed)
• Related to #16091

Original created by @u on 16128 (Redmine)