Re: [saag] draft-hallambaker-mesh-udf-02

[Michael Richardson <[e-mail]>]

    mcr>     Phill, I am reading your UDF document with the idea that IoT
    mcr> devices on constrained networks can send a reference to an IDevID
    mcr> rather than send the IDevID in-band.  This would apply to both DTLS and
    mcr> also to EDHOC and HIP (such as for 802.15.9 keying).

Phillip Hallam-Baker <[e-mail]> wrote:
    > If we went down that route, we might want to go full throttle and add
    > in a code point for a verbatim public key rather than a hash of a
    > public key. I have added that to the spec and removed it three times
    > because I do not make use of it in the Mesh.

I am not opposed to such a thing, but in my use case I would not use it.
I actually need the strong connection to the manufacturer, and the extensions
that PKIX provides. (Some new CoID would work equally well, but would also be
larger than a raw public key)

    > Another capability I left out is the authenticated only Locator
    > form. So the browser follows a link, gets back the result and compares
    > the fingerprint before accepting it. That could be a useful capability
    > for other applications. 

The fingerprint is in the link, along with the link?
That seems bigger, and does not help me.

    mcr>     You have specified rfc3986 rather than 3987, so your URI are not
    mcr> internationalized.  Is there a reason for that?

    > Lack of confidence in my knowledge of the subject material and lack of
    > time to get to a sufficient understanding.

    > Since the path-abempty of the URI is constrained to be base32 encoded
    > text, with punctuation (-/), the only place where there is scope for
    > internationalization is authority, that is the DNS name.

I am also not an internationalization expert.
We referenced iauthority in 3987 in draft-ietf-anima-bootstrap-keyinfra
rather than authothority from 3986.  This is within a PKIX extension.
Maybe we were wrong.

    > If we are forming a QR code, we might as well just use the punycode.

Well that eats up a 30% more bytes, doesn't it?  It's not like we have to put
it through an ASCII<->EBCDIC gateway like email.

    mcr>     Can you advance the use document without the rest of mmm?

    > Yes. I have presented the work as a single unit because this is what we
    > told the PeP people that we wanted when they came to SECDispatch. But I
    > am quite happy to go a-la-carte if people prefer.

    > The UDF doc is probably the one that is most easily separated from the
    > rest as it doesn't depend on any other part of the Mesh. It does depend
    > on draft-hallambaker-web-service-discovery but that is actually just a
    > profile of [RFC6763] and [RFC5785].

Can we do that, then?

-- 
]               Never tell me the odds!                 | ipv6 mesh networks [ 
]   Michael Richardson, Sandelman Software Works        | network architect  [ 
]     [e-mail]  http://www.sandelman.ca/        |   ruby on rails    [