6 ms·

PGP habits could be better

22 points by stargrave 7y ago

qrbLPHiKpiux 7y ago
PGP is dying a horrific death with the mass electronic and internet user. Since 1992, I’ve been in this game. It will never catch on.
- ryeights 7y ago
  That which never lived cannot die
tptacek 7y ago
What I'm really seeing here is a person documenting a whole lot of performative stuff they do --- grooming their keys, fastidiously using 4096-bit RSA (whatever breaks 2048 bit RSA is going to end RSA, by the way), signing other people's keys, creating lots of separate PGP keys and grooming the metadata on them, publishing a "proper canary" --- or wish they were doing. Which is a big problem with PGP. All the ceremony repels normal users, makes the system far harder to use, but also tightly binds enthusiasts to its subculture, making it that much harder to improve.
- sthatipamala 7y ago
  PGP seems like a leaky abstraction. I am highly technical and my eyes glaze over when I see terms like elliptic curve. What should I use if want to send dead simple, foolproof encrypted email? (Foolproof in that I can trust that updating the program will patch vulnerabilities and incorporate the state of the art methods)
  - tptacek 7y ago
    You can't send a foolproof encrypted email. Email is insecure and always will be. The idea that PGP could make it secure was always a fantasy. Efail should have been the final nail in the coffin. If you want to talk to other people securely, use a secure messenger; secure messenger cryptography is its own distinct subfield of cryptography research, and real secure messengers have cryptography purposely designed for it. It involves elliptic curves, but you don't have to think about that. Signal is the best of the lot, but Wire, or anything else that uses Signal Protocol is fine.
    - johnisgood 7y ago
      > Signal uses standard cellular mobile numbers as identifiers This is a big no. Even Wire uses metadata but only requires an e-mail account. I would like to see Ricochet (https://ricochet.im/ https://ricochet.im/) picked up by someone. Tor hidden services have been improved with added features which Ricochet does not utilize. I skimmed through the codebase and I came across some issues, heck, even the design is a bit obsolete as we have better algorithms with better parameters. Regardless, the idea is great and this is definitely doable, even by modifying Ricochet's codebase. There is Cwtch (https://cwtch.im/ https://cwtch.im/) that extends Ricochet to support multi-party messaging. I did not dig deep into it, but it may use the same algorithms. Additionally, it uses a language with a GC. I think a Ricochet-like IM software written in Ada/SPARK using libsparkcrypto (https://github.com/Componolit/libsparkcrypto https://github.com/Componolit/libsparkcrypto) would be ideal.
      - dngray 7y ago
        > Signal uses standard cellular mobile numbers as identifiers Exactly and what's to say I want some random guy on the internet being able to call my phone. Keep in mind tptacek quite literally pastes replies gushing about Signal/Wire in any thread related to PGP or instant messaging. They even paste it in threads which mention PGP but talk about other issues such as file encryption, signing etc, ie the uses not covered by an instant messaging platform. I have come to just skimming over anything they say because of it. Come on, put a bit of effort into your reply. > Wire uses metadata but only requires an e-mail account. It also doesn't offer federation. At this point I think Matrix is the best hope if you're okay with minimal metadata. It is also making huge strides of progress if you keep up to date with their development, so that's exciting. https://matrix.org/blog/posts https://matrix.org/blog/posts > There is Cwtch (https://cwtch.im/ https://cwtch.im/) that extends Ricochet to support multi-party messaging. I did not dig deep into it, but it may use the same algorithms. Additionally, it uses a language with a GC. I have high hopes for Cwtch. It seems to be a common thing people don't want to address: metadata. That said you can have a fairly meta-dataless experience with Matrix if you run a server and don't federate. Ideally I'd like to see this get solved https://github.com/vector-im/riot-web/issues/2320#issuecomment-485227466 https://github.com/vector-im/riot-web/issues/2320#issuecomme...
- drenvuk 7y ago
  So then we make a good PGP frontend which can properly utilize the web of trust. What a lot of smart people - you included - have noticed is that there are too many technical knobs to turn for doing certain actions like talking to someone or encrypting for someone while also being able verify and secure their communications or files. Even if PGP is used properly it still takes too many steps and too much know how. My take is that we should take out the labyrinthine intellectual masturbation that is key management, signing, encryption and relationship management and just let people use it for their activities. Whats more, saying that we shouldn't use PGP and instead should use separate tools for the job doesn't help either. Then we'd need to figure out how to use each individual tool for its purpose. That does nothing to empower users, it only swaps one pain for another. Keybase is almost there, Signal is almost there. But they're not what we really need.
johnisgood 7y ago
So what does it criticize exactly? PGP, OpenPGP, GPG, or...?
billpg 7y ago
Using-pink-in-moderation habits could be better.

22 points by stargrave 7y ago

PGP is dying a horrific death with the mass electronic and internet user. Since 1992, I’ve been in this game. It will never catch on.

ryeights 7y ago
That which never lived cannot die

tptacek 7y ago

What I'm really seeing here is a person documenting a whole lot of performative stuff they do --- grooming their keys, fastidiously using 4096-bit RSA (whatever breaks 2048 bit RSA is going to end RSA, by the way), signing other people's keys, creating lots of separate PGP keys and grooming the metadata on them, publishing a "proper canary" --- or wish they were doing. Which is a big problem with PGP. All the ceremony repels normal users, makes the system far harder to use, but also tightly binds enthusiasts to its subculture, making it that much harder to improve.

sthatipamala 7y ago
PGP seems like a leaky abstraction. I am highly technical and my eyes glaze over when I see terms like elliptic curve. What should I use if want to send dead simple, foolproof encrypted email? (Foolproof in that I can trust that updating the program will patch vulnerabilities and incorporate the state of the art methods)
- tptacek 7y ago
  You can't send a foolproof encrypted email. Email is insecure and always will be. The idea that PGP could make it secure was always a fantasy. Efail should have been the final nail in the coffin. If you want to talk to other people securely, use a secure messenger; secure messenger cryptography is its own distinct subfield of cryptography research, and real secure messengers have cryptography purposely designed for it. It involves elliptic curves, but you don't have to think about that. Signal is the best of the lot, but Wire, or anything else that uses Signal Protocol is fine.
  - johnisgood 7y ago
    > Signal uses standard cellular mobile numbers as identifiers This is a big no. Even Wire uses metadata but only requires an e-mail account. I would like to see Ricochet (https://ricochet.im/ https://ricochet.im/) picked up by someone. Tor hidden services have been improved with added features which Ricochet does not utilize. I skimmed through the codebase and I came across some issues, heck, even the design is a bit obsolete as we have better algorithms with better parameters. Regardless, the idea is great and this is definitely doable, even by modifying Ricochet's codebase. There is Cwtch (https://cwtch.im/ https://cwtch.im/) that extends Ricochet to support multi-party messaging. I did not dig deep into it, but it may use the same algorithms. Additionally, it uses a language with a GC. I think a Ricochet-like IM software written in Ada/SPARK using libsparkcrypto (https://github.com/Componolit/libsparkcrypto https://github.com/Componolit/libsparkcrypto) would be ideal.
    - dngray 7y ago
      > Signal uses standard cellular mobile numbers as identifiers Exactly and what's to say I want some random guy on the internet being able to call my phone. Keep in mind tptacek quite literally pastes replies gushing about Signal/Wire in any thread related to PGP or instant messaging. They even paste it in threads which mention PGP but talk about other issues such as file encryption, signing etc, ie the uses not covered by an instant messaging platform. I have come to just skimming over anything they say because of it. Come on, put a bit of effort into your reply. > Wire uses metadata but only requires an e-mail account. It also doesn't offer federation. At this point I think Matrix is the best hope if you're okay with minimal metadata. It is also making huge strides of progress if you keep up to date with their development, so that's exciting. https://matrix.org/blog/posts https://matrix.org/blog/posts > There is Cwtch (https://cwtch.im/ https://cwtch.im/) that extends Ricochet to support multi-party messaging. I did not dig deep into it, but it may use the same algorithms. Additionally, it uses a language with a GC. I have high hopes for Cwtch. It seems to be a common thing people don't want to address: metadata. That said you can have a fairly meta-dataless experience with Matrix if you run a server and don't federate. Ideally I'd like to see this get solved https://github.com/vector-im/riot-web/issues/2320#issuecomment-485227466 https://github.com/vector-im/riot-web/issues/2320#issuecomme...
drenvuk 7y ago
So then we make a good PGP frontend which can properly utilize the web of trust. What a lot of smart people - you included - have noticed is that there are too many technical knobs to turn for doing certain actions like talking to someone or encrypting for someone while also being able verify and secure their communications or files. Even if PGP is used properly it still takes too many steps and too much know how. My take is that we should take out the labyrinthine intellectual masturbation that is key management, signing, encryption and relationship management and just let people use it for their activities. Whats more, saying that we shouldn't use PGP and instead should use separate tools for the job doesn't help either. Then we'd need to figure out how to use each individual tool for its purpose. That does nothing to empower users, it only swaps one pain for another. Keybase is almost there, Signal is almost there. But they're not what we really need.

johnisgood 7y ago

So what does it criticize exactly? PGP, OpenPGP, GPG, or...?

billpg 7y ago

Using-pink-in-moderation habits could be better.