W3C home > Mailing lists > Public > public-linked-json@w3.org > October 2011

Re: ISSUE-31: Merge @type and @datatype / JSON-LD Telecon Minutesfor 2011-10-18

From: Gregg Kellogg <gregg@kellogg-assoc.com>
Date: Sun, 23 Oct 2011 15:23:23 -0400
To: Markus Lanthaler <markus.lanthaler@gmx.net>
CC: Linked JSON <public-linked-json@w3.org>
Message-ID: <E5964F1C-523A-4587-8CFC-CB216431611C@greggkellogg.net>
On Oct 22, 2011, at 8:12 AM, Markus Lanthaler wrote:

>> From reading the minutes and listening to the telecon recordings I think there have been quite some misunderstandings. Let me try to elaborate a bit on the consequences of merging @type and @datatype. 
> 
> @type can be used to set the type (think class) of a JSON object. For example you can use it to say that a specific JSON object in a JSON-LD document represents a person (foaf:person). On the other hand, @datatype, is used to specify the datatype of a JSON property. The only way to do that is when the extended form is used (in all other cases @coerce is used): 
> 
> {
>  ...
>  "property":
>  {
>    "@literal": "value",
>    "@datatype": "whatevertype"
>  }
> }
> 
> At the same time @type can't be used in such a construct as it would result in invalid triples. So having
> 
> {
>  ...
>  "property":
>  {
>    "@literal": "value",
>    "@type": "whatevertype"
>  }
> }
> 
> is invalid anyway. Thus, there is really no need to have two terms to set the datatype of a JSON object/property. In the telecon and on the mailing list some people argued that this will result in more invalid data. But I can't why this would be prevented by having two keywords? I think quite the opposite will happen, a lot of developers will probably use @type and @datatype in the wrong places if we have both of them.
> 
> Aliasing @datatype to @type as suggested on the call isn't a solution either. Processors would then have to check the surrounding tags to resolve the conflict - that's exactly the behavior we need when we merge the two. So either we disallow aliasing @datatype to @type or we just move on and merge them.
> 
> The only thing developers could get wrong is to use a class type when they mean a datatype or vice versa. But that can't be prevented by keeping @datatype and @type separate. I can't see with the best will in the world one argument to keep both of them. But perhaps someone will teach me better.

Well, I guess I'm in serious jeopardy of waffling on this issue (good thing I'm not planning a run for office). I do think that eliminating an unnecessary keyword is a good idea. We also never really discussed replacing @subject with @iri, which is quite similar to this one.

The probably with the teleconferences as a decision making vehicle is that not everyone that's interested can participate. We're probably better to do straw polls on the mailing list, or GitHub issues list and use the consensus from those forums for making such decisions.

Gregg

> --
> Markus Lanthaler
> @markuslanthaler
> 
> 
> 
> 
Received on Sunday, 23 October 2011 19:24:15 GMT

This archive was generated by hypermail 2.3.1 : Tuesday, 26 March 2013 16:25:35 GMT