I'm still think that OpenPGP is the best solution for the first two. But the two tools at the top have the same problems as pgp.

How do you trust some key material. The WoT is a complete failure, but modern solutions like WKD are amazing and make PGP just plain work.

The one nice thing about age is the ability to encrypt to a GitHub users key. However, even that is possible with OpenPGP. Here is a tool that I made ~7 years ago that does something similar: https://github.com/georgyo/sshcrypt

Another thing minisign/age don't have is a method to say that a key is compromised.

I don't think the solution to the problems with OpenPGP is too just ignore the problems and switch to using bare keys like new tools are doing.

I wouldn’t hastily recommend unproven tools in the area security.

Here is an example of a cool tool with modern cryptography, forward secret etc, often recommended in HN as an alternative to Wormhole:

https://redrocket.club/posts/croc/

It turned out that plaintext could easily be recovered! One mistake and 100% broken.

There are benefits to an industry standard protocol.

croc wasn't recommended by people on HN who knew something about cryptography, and they actually warned against it.

They recommended Brian Warner's magic-wormhole, Signal, Tarsnap, age, Signify, Minisign, libsodium.

Signify and Age are not unproven.

Signal is an instant messenger ... which is fine, but sometimes you really do need the extra security provided by an offline messaging solution like OpenPGP. The Cellebrite thing was a good example of this. Cellebrite doesn't get OpenPGP messages as they can and are usually protected by a passphrase. Cellebrite gets Signal messages.

In that case I would recommend Age. It’s a modern and sound replacement for GPG.

https://age-encryption.org/

Not really. My opinion in some detail:

* https://articles.59.ca/doku.php?id=pgpfan:agevspgp

Age might be such a replacement some day for just the encryption function. But it has a ways to go. Actually defining the format would be a good first step.

For encrypted backups, I believe that Restic is a better tool https://restic.net/

Enabling 'disappearing messages' will help I think.

Don’t forget, keys.pub for signing, encrypted messaging, and has wormhole in it

Age is not even close to being some sort of GPG replacement:

* https://articles.59.ca/doku.php?id=pgpfan:agevspgp

I don't think that's a fair comparison. Age was never meant to fully replace GPG, so complaining about missing features it was never meant to have doesn't make sense.

The only valid criticism is about its handling of corrupted data, but for that too you can use an external tool like PAR2 to generate recovery data. I do this with my GPG backups and other data as well.

I like that age follows the Unix "do one thing well" philosophy, and that I can use other similarly scoped tools for other features. There are still some things I'm missing, but these are slowly being worked on[1,2].

[1]: https://github.com/FiloSottile/age/issues/137

[2]: https://github.com/FiloSottile/age/pull/132

I am not sure how PAR2 helps here if your encryption utility refuses to return any data in the first place.

I was responding to a comment that said that age was a "standard replacement" for GPG.

Signal is centralized and in no way can replace emails.

Email is default plaintext, can't force encryption on groups, leaks --- in fact, gleefully publishes --- the most important dragnet surveillance metadata (even when ostensibly encrypted), creates long term archives of secrets, is built on long term secrets, and, to boot, is in the real world itself effectively centralized. It's hard to even imagine a secure messaging system less safe than email. Recommending encrypted email is simply malpractice.

Recommending signal fails to realize that it can be compromised in many ways precisely because it is centralized. A false sense of security.

Email clients and server instances are just as centralized in practice. Yet email servers almost always know everything about you in contrast to Signal. And you can prove this by reading their code and by the fact that they respond to subpoenas with as little info as they do.

How would signal be compromised that email could not and which signal is not better prepared for?

> Email clients and server instances are just as centralized in practice.

I don't follow. Not everyone uses Gmail or the big email providers.

If anyone on the communication thread uses GMail or the other big providers then the entire conversation is subject to the weaknesses of the protocol.

Right now that sits at about a 1 in 5 chance for Gmail alone.