Tim W is a user on cantos.social. You can follow them or interact with them if you have an account anywhere in the fediverse.
Tim W @tw

I'm bored. Ask me questions about , , , , or anything else, I'll be around the next ~2 hours!

(More topic ideas in my pinned posts, probably.)

@tw why do people still think DNSSEC has any tangible value. It’s garbage.

@feld I would argue it does in fact do what it says it does, it's just... extremely difficult to get it to do that correctly, perhaps?

@tw unless every client device on the planet runs their own DNS recursor with DNSSEC validation it provides very little value in the real world

@feld DNSSEC is one area of DNS I never dived very deeply into; that said, it is my understanding (wave hands, wave hands) that a validating client (which doesn't have to be a full recursive resolver) still gains protections from a signed zone.

@feld I definitely do agree that the implementation / trying to actually deal with it is garbage though.

@tw I think it’s the wrong place in the stack to try to do authentication. Most DNS attacks are going to be against clients / browsers so as long as you’re using HTTPS you’re fine. The difficulty in forging TLS today is extremely high.

I much prefer dnscrypt which gives me a protected path to a recursor I can trust has strong security controls and monitoring for tampering / poisoning.

@feld dnscrypt is really complementary to DNSSEC in that way - it's preventing the MITM between you and the resolver, and DNSSEC is preventing MITM between the resolver and whatever you're looking up.

@tw kind of, but it doesn’t work that way in practice. The client using the recursor through dnscrypt still has to request DNSSEC records or they won’t be used and verified. So no good for most use cases, especially mobile clients.

@feld "as long as you're using HTTPS you're fine" presupposes a non-broken certificate authority system, unfortunately.

@tw it’s not broken anymore with Certificate Transparency