Re: tl26 pretest start

Hi Norbert,

On Mon, 2026-02-23 at 19:05 +0100, Norbert Preining wrote:
> If you have better suggestions, I am more than happy to discuss them,
> but as far as I see there is no way to have cryptographic check **AND**
> a kind of "web of trust" to ensure that nothing has been tampered.
>
> There are a lot of other and better tools for signing **ASSUMING** that
> we can guarantee the existence of a untempered public key. But this
> cannot be guaranteed due to the distribution network, and the only
> guarantee is a web of trust.

What's wrong with the following?

    $ wget https://tug.org/texlive/public-key.txt

wget checks the TLS certificates, so that ensures that the key really
was downloaded from tug.org. Now, you could argue that that doesn't
"really" ensure that the file came from TUG, but between certificate
transparency and CAA records, we (TUG) can ensure that only the tug.org
server is able to obtain a proper certificate.

You could also argue that someone could have hijacked the tug.org TLS
private key, but considering that tug.org also holds the gpg private key
(needed for signing package updates), this argument applies equally to
gpg.

In fact, I would actually argue that using the TLS certificates to
validate a public key is *more* secure than the GPG web of
trust---everyone already has the public keys for the TLS roots on their
system, while much fewer people have any GPG keys in the WoT, so people
are way more likely to be able to correctly validate a key obtained over
HTTPS than one obtained from a gpg keyserver. Plus, TLS certificates are
replaced every 90 days, while our gpg key is over 10 years old, so the
odds of our gpg private key having ever been exposed is greater than the
odds that a current TLS private key has been exposed.

Thanks,
-- Max