Re: [Dcrup] Hashed Key Records

[denis bider <[e-mail]>]

I propose that the best fingerprint encoding designed so far is the
Bubble-Babble encoding by Antti Huima. Example:

xedec-nypul-taboz-mugaz-supak-nekyc-bibec-busek-fegon-gekug-loxix

It is easy to compare words both visually, and on the phone. Users like it.

I found the spec preserved here:

http://wiki.yak.net/589/Bubble_Babble_Encoding.txt

It was intended to be standardized as part of SSH, but somehow didn't make
it to RFC.

A number of implementations implemented Bubble Babble, and still have it.
However, open source implementations (you know who you are!) went with
hexadecimal (MD5) and now Base64 (SHA-256), and now we're stuck with that.

If we want to standardize a universal fingerprint format, Base32 is
certainly better than either hex or Base64. However, I'd suggest checking
out Bubble Babble.

denis



On Thu, Jun 22, 2017 at 5:44 PM, Phillip Hallam-Baker <[e-mail]
> wrote:

> I would really like to get to a point where we could use one fingerprint
> format everywhere. This would then allow fingerprints to be cut and pasted
> from one place to another.
>
> When I suggest we do something of this sort, people say 'write a draft'.
> Once I write a draft they say I am peddling my proposal. So pretend that I
> don't have a draft yet. These are the features I think are needed:
>
> 1) Use Base32 encoding.
>
> DNS labels are case insensitive. Using fingerprints in DNS labels is very
> powerful. It allows DANE like effects without the need for DANE records or
> DNSSEC.
>
> 2) Incorporate the algorithm identifier into the fingerprint, do so in a
> way that ensures a Base32 fingerprint will never be confused with a PGP
> fingerprint.
>
> 3) Include a description of the content type in the target of the
> fingerprint. This prevents semantic substitution attacks.
>
> 4) Allow for optional grouping of characters to encourage readability.
>
> On Thu, Jun 22, 2017 at 7:17 PM, Scott Kitterman <[e-mail]>
> wrote:
>
>> It seemed like the hashed RSA variant ought to be easy enough to
>> implement, so
>> I decided to give it a go.  I almost immediately ran into a question.  For
>> terminology, I'm using what openssl says [1]
>>
>> I am assuming that "SHA-256 hash of the public key, stored in base64"
>> means
>> use the output of the digest in binary form (--binary) option and encode
>> it in
>> base 64.  Is that correct?
>>
>>
>> So if my signing key were (I know, but just to make sure I understand the
>> tranformations):
>>
>> 'Hello World'
>>
>> Then the the p tag in DNS would be:
>>
>> p=pZGm1Av0IEBKARczz7exkNYsZb8LzaMrV7J32a2fFG4=
>>
>> Am I understanding it correctly?
>>
>> A python one liner:
>>
>> >>> base64.b64encode(hashlib.sha256(b'Hello World').digest())
>> 'pZGm1Av0IEBKARczz7exkNYsZb8LzaMrV7J32a2fFG4='
>>
>> Scott K
>>
>>
>>
>> [1] https://www.openssl.org/docs/man1.1.0/apps/sha256.html
>>
>> _______________________________________________
>> Dcrup mailing list
>> [e-mail]
>> https://www.ietf.org/mailman/listinfo/dcrup
>>
>
>
> _______________________________________________
> Dcrup mailing list
> [e-mail]
> https://www.ietf.org/mailman/listinfo/dcrup
>
>