RE: Basic Roles Proposal

Hi Rob,

Thanks for the details.

Let me confirm. If a person/system would like to set language to the body text for some reasons, the person/system has to take 3.1.2 or 3.1.3 approach and add unknown role property. Correct?

I don’t know how many annotations would be provided without roles, and I don’t know how it will be useful, but I’m afraid that annotation systems might need to add the unknown role frequently.
As Youtube comment list shows, we will see annotation texts written in variety of languages in the same page. I won’t see any issues as long as the texts are written in Western languages, but Chinese/Japanese/Korean text could be a problem, because it needs each language font to render the text, due to Unicode CJK unified ideographs issue. Simply said, slightly different shape of glyphs in each language are assigned to the same Unicode code point. As an English text which shows “alpha”, “beta”, *gamma”, instead of “A”, “B”, “C” makes difficult to read, Chinese/Japanese/Korean users don’t like to read text rendered in different language fonts. Then, cutting of the language capability is an issue.

In case 3.1.8 provides huge benefit to users and implementers, adding explicit explanation would be good, but if it does not, I will appreciate if removing the item itself would be taken into consideration.

Takeshi Kanai

From: Robert Sanderson []
Sent: Friday, August 28, 2015 12:58 AM
To: Kanai, Takeshi
Cc: Web Annotation
Subject: Re: Basic Roles Proposal

Hi Takeshi,

On Thu, Aug 27, 2015 at 12:53 AM, Kanai, Takeshi <<>> wrote:
 3.1.8 Literal Bodies Cannot Have Roles
Can I think “language” in this description points to dc:language which is delivered from @context, and it is still possible to set @language to body?
I would like to make sure whether the simple literal body prohibits even having native properties in each serialization format, such as @language in JSON-LD and “LANGTAG” (@en, for example) in Turtle, or not.

In JSON-LD it is technically possible to use @language with a literal.  In RDF, this translates to "some string"@en.  Also @type to set the format:  "<span>some string</span>"^^rdf:HTML

The issue is that it's not possible to do both at once and have both a language and a format associated with a literal.  Thus, to avoid the inevitable mistakes between @language and dc:language, @value and rdf:value, the decision was to only allow literals that were plain text strings and did not have a language associated with them.

The role proposal doesn't change that either way (it wasn't possible before and still isn't), but no it's not possible to set @language on a literal body.  This is not explicit in the current model documentation, and should be.

 3.1.7 Multiple Targets with Roles Example
I think this is an editorial issue. I can see ”lit” prefixes in the examples, but what it stands for is not described. Will it be fixed?

I can fix that :)

I'll add some descriptive text before the examples, and a second context that defines the lit prefix.

Many thanks Takeshi!


Received on Friday, 28 August 2015 11:08:34 UTC