Re: [T2TRG] draft-hartke-t2trg-ciri-00 review

[Klaus Hartke <[e-mail]>]

Hi Dave,

thanks for bringing up this interesting point.

The draft essentially describes a CBOR variant of the way CoAP
serializes the Request URI: The URI is split into its components (+
the path component into its segments) and each component becomes one
"option". In these options, no percent encoding is performed. So, as I
see it, the following are all serializations of the same data:

URI:
   coap://example.com:5683/city/Montr%C3%A9al

IRI:
   coap://example.com:5683/city/Montréal

CoAP:
   Uri-Host: "example.com"
   Uri-Port: 5683
   Uri-Path: "city"
   Uri-Path: "Montréal"

CIRI:
   [1, "coap",
    2, "example.com",
    4, 5683,
    6, "city",
    6, "Montréal"
   ]

I don't think the problems of one serialization have much impact on
the standardization of a new serialization.

(What's the name of the data model of which URI, IRI, CoAP and CIRI
are serializations?)

Klaus

On Mon, 21 Jan 2019 at 21:11, Dave Thaler
<dthaler=[e-mail]> wrote:
>
> FYI, the IRI WG intended to update RFC 3987 and shut down because it could not reach
> consensus on IRIs.  In particular, RFC 3987 does not match what major implementations
> (i.e. browsers) actually do, and so there's a major incompatibility between current practice
> and what RFC 3987 does.  There does exist at least one RFC 3987 implementation out there
> in Windows WinRT APIs though, but I don't know if there's others.   In practice, the
> recommendation from the internationalization community in the IETF tends to be to not
> use IRIs.   That is, protocols should only ever put URIs (not IRIs) on the wire, and conversion
> from internationalized strings to/from URIs is done at the UI layer.
>
> With that in mind, the case for using IRIs (as opposed to URIs) in Constrained Restful
> Environments seems to not have much real-world applicability since it goes against
> the internationalization guidance.
>
> It's fine for a research project, which T2TRG is allowed to do, but may not be very useful
> in practice, from what I've seen.
>
> Dave
>
> P.S.: An example of one incompatibility between RFC 3987 and what is widely implemented
> in browsers, is how percent encoding is interpreted.  In RFC 3987 it's percent-encoded UTF-8,
> whereas in implementations it's percent-encoded in whatever the charset is of the document
> (or whatever) you got the IRI out of, which may or may not be UTF-8.   Again this is just one
> of many examples.