DNS-Based Authentication of Named Entities (DANE) Bindings for OpenPGP | Hacker News Reader

DNS-Based Authentication of Named Entities (DANE) Bindings for OpenPGP

rfc-editor.org

48 pointsokket11 comments

Link with more HTML: https://tools.ietf.org/html/rfc7929

And here with timeline etc: https://datatracker.ietf.org/doc/rfc7929/

All docs from the DANE working group: https://datatracker.ietf.org/wg/dane/documents/

I'm still annoyed that browsers didn't start supporting DANE

Browser developers (and by extension, web standards developers) appear to be extremely reluctant to work at any protocol level other than HTTP.

To this aversion I attribute several pathological outcomes that affect many of us. Layering-violation constructs like WebSockets, failure to use SRV leading to hacked-up apex DNS A record tracking, timeouts/misbehaviours giving IPv6 a bad name, the general reinforcement of HTTP as an inefficient and poorly specified RPC transport.

And, yes, ignoring DANE. Fun fact: even using DNS A records for endpoint name resolution in HTTP is a convention; it is not mandated in the document body or in any of the normative references.

My greatest disappointment in HTTP/2 is that it reinvents multiplexing. Not so much that it wasn't necessary (it was, at least it was if you are Google or Facebook or Twitter scale), but that it occurs inside the protocol rather than in a separate layer that other future protocols could use. To me, that's exemplary of how HTTP has been developed despite, rather than in concert with, other protocols.

One of the main implementors of TLS in Chrome answered why: <https://www.imperialviolet.org/2015/01/17/notdane.html>.

It boils down to:

* Getting the DANE and DNSSEC records to the end users' machines is surprisingly difficult (they recorded 4-5% failure to lookup TXT records that they knew existed), so requiring that it be done to accept a certificate isn't going to work.

* DNSSEC uses 1024-bit RSA, which is now considered too weak to use in web browsers, and the IETF/ICANN refuse to up its security.

> DNSSEC uses 1024-bit RSA, which is now considered too weak to use in web browsers, and the IETF/ICANN refuse to up its security.

False. The root KSK is currently 2048 bits¹, and the ZSK will be changed this year from 1024 to 2048 bits².

① https://www.iana.org/dnssec/icann-dps.txt

② https://www.icann.org/en/system/files/files/ksk-rollover-ope...

I agree, but there aren't a lot of websites that implemented DANE or even DNSSEC. You can install this add-on to validate the records: https://addons.mozilla.org/nl/firefox/addon/dnssec-validator...

It's chicken and egg though. If browsers started supporting it more sites would start using it and vice versa.

How is this going to make PGP easier for the average user?

This protocol can provide an alternative to the web of trust. The verification method of web of trust can't be easily automated, this protocol can. For example with this protocol an email client can lookup the matching PGP-key for an email address. This could make everyday use easier, but people will have to setup PGP, generate keys and then publish them. In my opinion this is still too difficult for the average user.

I's not. The majority of email users don't even know what PGP is.

Reply on news.ycombinator.com