AW: thing descriptions

Hi Dave, Sebastian, all,

I agree it is desirable for restricted devices to provide "shortcuts" for constrained devices. That said, I don't think we should allow “boolean” instead of “xsd:boolean" neither “uint32” instead of “xsd:unsignedInt”.

Please let me state the reasons.

First of all, introducing new names may shorten the message but on the other hand they also introduce more complexity. What if I still want to use “xsd:unsignedInt”? Is this forbidden, allowed, acceptable?

Second, if it is about message size, reducing the stream by some bytes only tackles the issue just half way.

Third, if we want facilitating highly resource constrained devices being part of the system we might consider using identifiers instead of strings. It is not only shorter but in addition it is also easy to switch according to identifiers compared to do string comparisons.

Fourth, I believe we should consider existing techniques offering those afore mentioned "shortcuts" for constrained devices instead of inventing it from scratch. One solution could be W3Cs EXI format but there could be more relevant techniques...

Hope this helps,

-- Daniel







________________________________

Von: Kaebisch, Sebastian [sebastian.kaebisch@siemens.com]
Gesendet: Freitag, 17. Juli 2015 10:52
An: Dave Raggett
Cc: public-wot-ig@w3.org
Betreff: AW: thing descriptions

Hi Dave,

that’s a good point. On the other hand we can just simple rely on an existing standard. I think, this is also an interesting discussion for the breakout session in 2 weeks.

Best wishes
Sebastian

Von: Dave Raggett [mailto:dsr@w3.org]
Gesendet: Mittwoch, 15. Juli 2015 20:44
An: Kaebisch, Sebastian
Cc: public-wot-ig@w3.org
Betreff: Re: thing descriptions


On 15 Jul 2015, at 17:11, Kaebisch, Sebastian <sebastian.kaebisch@siemens.com<mailto:sebastian.kaebisch@siemens.com>> wrote:

typically, RDF data relies on a type system coming from XSD declarations. Using this, we can declare and use data types which are mainly relevant for highly resource constrained devices such as byte, short, enums, etc..

Hi Sebastian,

Right, and we can take advantage of this. However, it will be worth looking for a way to minimise the overheads in simple data models. Can we avoid having to give namespace prefixes by using the default context to bind common terms to the full URI, e.g. can we say “boolean” rather than “xsd:boolean".  I believe that it will be very important to listen to web developers, who want simple notations and who in the past have shown low tolerance for name space prefixes (e.g. in HTML). Another issue is the verbosity of XSD data types, e.g. “unsignedInt” compared to “uint32”.  Long strings are costly for constrained devices and protocols with short packet sizes.

Best regards,
—
   Dave Raggett <dsr@w3.org<mailto:dsr@w3.org>>

Received on Monday, 20 July 2015 09:06:24 UTC