Re: [saag] [Trans] draft-iab-crypto-alg-agility-00

On 9 April 2014 01:49, Phillip Hallam-Baker <hallam@gmail.com> wrote:
> Since TRANS is joined to X.509 at the hip, how about we just shovel
> all the metadata describing the configuration of the notary into the
> certificate that signs the notary log outputs periodically?

I guess if you are a CA everything looks like a certificate.
Currently, logs do not use certificates to manage their own keys. I
guess they could. But I thought you preferred JSON for new stuff?

> I am assuming here that the signing hierarchy for the log has an
> offline key that periodically signs the online key.

Again, currently, no. Seems to me that you can introduce a lot of
complication this way not needed for the only known client use case
(deployment in Chrome). Not against specifying it, to be clear, but
unsure it should live in the same doc. Or hold it up.

> The cert for the
> online key can specify the notary algorithm by OID (its ASN.1 after
> all) and any other parameters that might be useful. Probably want to
> allow for different accumulation strategies in case we ever decide a
> binary tree isn't the only choice. Might have some other options.
> Might want to be able to specify the CPS/CP equivalents for the
> notary, etc.
>
>
>
> On Tue, Apr 8, 2014 at 10:28 AM, Dmitry Belyavsky <beldmit@gmail.com> wrote:
>> Hello Ben,
>>
>>
>> On Tue, Apr 8, 2014 at 6:21 PM, Ben Laurie <benl@google.com> wrote:
>>>
>>> On 8 April 2014 15:18, Salz, Rich <rsalz@akamai.com> wrote:
>>> >> > I do not understand why metadata is more secure then the data itself.
>>> >
>>> >> It is created by a different authority.
>>> >
>>> > ?  Is this in the part of the RFC that is still TBD?
>>>
>>> The RFC describes how logs work and how clients work. It does not
>>> describe how clients decide what logs they are prepared to accept. I
>>> am not sure it should.
>>>
>>> But whoever does also decides whether the algorithms in use by the
>>> logs are acceptable and tells the client what those algorithms are
>>> (along with other things, like the log's key, base URL and MMD).
>>>
>> I think that the client should be able to find out the algorithm used by log
>> because it cant'be changed during the log lifetime. And if the RFC specifies
>> the URIs for certificate submit, it seems to me that it's reasonable to
>> specify the URI for finding out the algorithm. But I prefer to leave out of
>> band of the protocol only the data that can't be passed using it.
>>
>> Thank you!
>>
>> --
>> SY, Dmitry Belyavsky
>>
>> _______________________________________________
>> Trans mailing list
>> Trans@ietf.org
>> https://www.ietf.org/mailman/listinfo/trans
>>
>
>
>
> --
> Website: http://hallambaker.com/

-- 
Certificate Transparency is hiring! Let me know if you're interested.