Re: [secdir] dir review of draft-laurie-pki-sunlight-05

[Tobias Gondrom <[e-mail]>]

On 23/02/13 00:36, Ben Laurie wrote:
> On 10 February 2013 02:54, Tobias Gondrom <[e-mail]> wrote:
>> On 10/02/13 04:47, Ben Laurie wrote:
>>> On 9 February 2013 13:34, Tobias Gondrom <[e-mail]> wrote:
>>>> Hi Ben,
>>>>
>>>> I also just read through your draft in version -07.
>>>> I can see the draft consists of two parts:
>>>> 1. data structure
>>>> 2. protocol.
>>>>
>>>> For part #1 the data structure: in case you are not aware of it, some
>>>> years ago the IETF LTANS WG has done something a bit similar in a more
>>>> generic way (i.e. for any data not only for certificates) in form of
>>>> RFC4998 and RFC6283
>>> Interesting. I was not aware of these. From a quick skim they are
>>> indeed similar, but would need a bunch of added machinery to get them
>>> to where CT is (e.g. not append only, no concept of MMD).
>> You are welcome.
>> I believe the gap is mostly towards the protocol side (e.g. including
>> MMD). As the RFCs only define the data structure.
> It seems like a lot of work to inherit an essentially trivial data structure.

Well. Whether you like to take it or parts of it or re-write/re-invent
the wheel is obviously up to you.
(and I agree that the base data structure (Merkle Hash Tree) itself is
pretty trivial. It gets complicated if/when you need to switch to new
algorithms over time, which I believe is ERS has proven very useful in
implementations. Whether that might be a potential use case for your
system, shall be for others to judge.)
I only thought I should make you aware of the existing RFCs.


>
>>>> with a number of implementations by major ECM and
>>>> DMS vendors.
>>> No idea what ECM or DMS are in this context.
>> ECM: Enterprise Content Management
>> DMS: Document Management System
>> (Systems that store electronic documents/data.
>> And protect the proof of integrity / non-repudiation with Timestamps and
>> RFC4998/RFC6283.)
> These systems rely on trust. :-)
>
>>
>>>> Just as a thought, maybe helpful looking at or even for re-use instead
>>>> of re-inventing the wheel?
>>>>
>>>> Best regards, Tobias
>>>>
>>>>
>>>>
>>>> On 30/01/13 18:15, Ben Laurie wrote:
>>>>> On 29 January 2013 21:28, Jeffrey Hutzelman <[e-mail]> wrote:
>>>>>> On Tue, 2013-01-29 at 11:35 +0000, Ben Laurie wrote:
>>>>>>> On 24 January 2013 19:06, Jeffrey Hutzelman <[e-mail]> wrote:
>>>>>>>> Similarly, as an anti-spam measure, this document proposes that logs accept
>>>>>>>> only certificates which chain back to a known CA, and requires that logs
>>>>>>>> validate each submitted certificate before appending it to the log.  This
>>>>>>>> sounds good, but it's not the only possible mechanism, and so I think MUST
>>>>>>>> is too strong here.  Additionally, there is no discussion of the security
>>>>>>>> implications if a client depends on a log to do this and the log does not
>>>>>>>> actually do so.  Rather than requiring that logs validate every submitted
>>>>>>>> certificate, the document should only RECOMMEND that they do so, and make
>>>>>>>> clear that clients MUST NOT depend on such validation having been done.
>>>>>>> On second thoughts, whilst that is an effective anti-spam measure, it
>>>>>>> is also part of the functionality of CT: i.e. to identify misissue and
>>>>>>> give some means to do something about it. The CA check ensures we have
>>>>>>> someone to blame for misissue.
>>>>>> Hrm.  I sort of thought the idea was for the logs to be untrusted
>>>>>> repositories, able to be audited but not themselves expected to detect
>>>>>> problems.  If logs are expected to do validation of this sort, is there
>>>>>> a way for a third party to discover whether they are doing so (or at
>>>>>> least, whether they are accepting certificates they shouldn't)?
>>>>> A third party can indeed verify this - they just watch the log like
>>>>> any monitor does.
>>>>>
>>>>>>> I am not averse to suggestions that achieve the overall aim, but I
>>>>>>> don't see the virtue of leaving it vague in the description of the
>>>>>>> experiment we are actually running.
>>>>>> I'm not suggesting vagueness; rather, I'm merely suggesting downgrading
>>>>>> a MUST to a SHOULD, which is still quite strong.  What happens if
>>>>>> someone wants to start logging certs issued by a private CA, or
>>>>>> self-signed certs they have observed, or...?
>>>>> I don't see an issue with logging certs from a private CA. As for
>>>>> self-signed certs, I don't see the point, but I guess if someone
>>>>> figures out a point we can relax it in the next version.
>>>>>
>>>>>> I'm suppose I'm OK with keeping the scope narrower than that for
>>>>>> purposes of the experiment, as long as it is possible to relax the
>>>>>> requirement later without breaking the system.  Hence the importance of
>>>>>> making it clear that clients must not rely on logs to have done
>>>>>> validation (on which point I think we've already reached agreement).
>>>>> Cool.
>>>>> _______________________________________________
>>>>> secdir mailing list
>>>>> [e-mail]
>>>>> https://www.ietf.org/mailman/listinfo/secdir
>>>>> wiki: http://tools.ietf.org/area/sec/trac/wiki/SecDirReview