Re: [TLS] ESNI PEM files

[Stephen Farrell <[e-mail]>]

On 25/10/2019 14:45, Eric Rescorla wrote:
> Are you suggesting that these strings appear on the wire in DNS? 

Nope.

> Or is this
> just an internal implementation covnenience?

A bit more than that, but yes. I have a (temporary) tool for
generating these files and then need to import the output from
that into e.g. nginx or lighttpd. Some applications support
more than one TLS library, so having a well defined format
seems like it'd help with ESNI deployment.

Basically this is the same logic that lead to pkcs8 PrivateKey
being worth documenting in an RFC.

Cheers,
S.

> 
> -Ekr
> 
> 
> On Fri, Oct 25, 2019 at 3:55 AM Stephen Farrell <[e-mail]>
> wrote:
> 
>>
>> Hiya,
>>
>> To date, my ESNI code [1] has been dealing with files containing
>> the binary encoding of ESNIKeys and a separate PEM file containing
>> the related private key.
>>
>> As part of getting nginx to work with ESNI [2], it'd be easier
>> for me to deal with a single file containing both private key
>> and the ESNIKeys (or ESNIConfig) value. (One of the upstream
>> maintainers didn't like how I handled configuring ESNI, so the
>> change is really to make him happier, but I think it's likely
>> a generally useful thing:-)
>>
>> Since mixing binary and PEM encoding in one file would be ickky,
>> it'd be nice if we had a PEM file convention for the public keys
>> used in ESNI. And of course, it'd be much nicer if everyone did
>> the same thing here.
>>
>> What I'm coding up now would handle files like:
>>
>> "
>> -----BEGIN ESNIKEY-----
>>
>> /wGeNLTKACQAHQAg/JLBUtPE5wp4CU2bbTihSPeP3113kz1J80X/Iy1Y2EYAAhMBAQQAAAAAXO6j
>> /gAAAABc995+AAA=
>> -----END ENSIKEY-----
>> "
>>
>> ...where the content is (unsurprisingly:-) a base64 encoded ESNIKeys.
>>
>> And if both the private and public components are in one file,
>> it'd look like:
>>
>> "
>> ----BEGIN PRIVATE KEY-----
>> MC4CAQAwBQYDK2VuBCIEICAHxXknil9tI2qZ+USRouNwXp0LxlUB85l0/xbhZ4Va
>> -----END PRIVATE KEY-----
>> -----BEGIN ESNIKEY-----
>>
>> /wGeNLTKACQAHQAg/JLBUtPE5wp4CU2bbTihSPeP3113kz1J80X/Iy1Y2EYAAhMBAQQAAAAAXO6j
>> /gAAAABc995+AAA=
>> -----END ENSIKEY-----
>> "
>>
>> I'd be happy to change that however increases the probability
>> that other code bases do the same thing.
>>
>> If this is useful, I could do up a PR with a bit of text saying
>> to use "ESNIKEY" (or whatever) as the label when storing these
>> things in PEM files. Given RFC7468 documents other PEM file
>> things, it seems reasonable to do the same for ESNI.
>>
>> Cheers,
>> S.
>>
>> [1] https://github.com/sftcd/openssl/tree/master/esnistuff
>> [2] https://github.com/sftcd/nginx
>> _______________________________________________
>> TLS mailing list
>> [e-mail]
>> https://www.ietf.org/mailman/listinfo/tls
>>
>