Re: [xri] DNS XRD Discovery proposal

[Ben Laurie <benl@google.com>]

On Wed, Dec 17, 2008 at 1:52 PM, Peter Davis <peter.davis@neustar.biz> wrote:
> On Dec 17, 2008, at 5:11 AM, Ben Laurie wrote:
>
>> On Tue, Dec 16, 2008 at 9:07 PM, Peter Davis <peter.davis@neustar.biz>
>> wrote:
>>>
>>> I've finally gotten around to adding to the wiki my suggestions on the
>>> use
>>> of the DNS for XRD discovery [1].
>>>
>>> I did not get all the example URIs in that i wanted, but thought, given
>>> recent conversation, i should at least demonstrate how DNS would provide
>>> XRD
>>> (and site-meta), and perhaps simplify the XRD location processing a bit.
>>
>> Not sure why this mentions SMS? Other than that, looks good to me,
>> though a couple of points:
>
> I clarified in my reply to Brian's similar question.  I will work in an SMS
> and a few other non-browser examples over time.
>
>> 1. Under "Services" you say you will define 1 service, and then go on
>> to define 2.
>
> I added one mid-writing and forgot to update that sentence.  thanks.
>
>> 2. The Service is "U2MD+xrd" (for example), but RFC 3403 has them the
>> other way round: "xrd+U2MD".
>
> Actually, the DNS database specified in 3403 says only:
>
>        SERVICES
>        A <character-string> that specifies the Service Parameters applicable
> to
>        this this delegation path. It is up to the Application Specification
> to specify the values found in this field.
>
> the example uses service identifiers before the transform hint (U2MD) in
> 3403.  the ABNF i supplied is pretty much a clone of RFC3761 (ENUM), so the
> pattern sequencing is their fault :-P

I didn't mean to suggest it was wrong, just different from their
examples for no apparent reason.

>> 3. The examples are perhaps a little too lax, since
>> http://foo.example.biz/any/url/at/all?abc=def#blah would transform to
>> https://bar.example.biz/any/url/at/all?abc=def/user-xrd#blah, which
>> doesn't seem right.
>
> I hinted at that in the draft.  Of course the regex is up to individual
> implementations, so if query params and fragment identifiers are
> undesirable, NAPTR publishers would accommodate for that.

Of course, but some level of realism might be nice.

>> 4. Not clear to me why the #blah gets preserved?
>
> I showed that one deliberately, since it seemed important for some for key
> selection and trust processing discussions rooted in [1]

I'm sure it was deliberate, but I still don't understand why it gets preserved.

>
> =peterd
>
> [1]
> http://www.oasis-open.org/apps/org/workgroup/xri/email/archives/200812/msg00082.html
>