Re: RFC: XDM RDF schema

On 7/15/24 21:33, Liam R. E. Quin wrote:
> On Mon, 2024-07-15 at 21:18 -0500, Paul Tyson wrote:
>> This could be considered a long-overdue update of the original RDF
>> schema for the XML Information Set [2][3].
>> One use case is to standardize a graph transformation path from RDF
>> to (non-RDF) XML.
> I am not at all sure that there’s a single useful XML representation of
> any given RDF graph, let alone of RDF graphs in general.
Yes, exactly. One might want to go to XHTML, DITA, DocBook, XSL-FO, SVG, 
or whatever. And within those, any particular subgraph might be best 
rendered as a table, list, divisions, boxes and arrows, etc. All of 
which would be specified in the RDF-to-XDM transformation. The 
XDM-to-XML transformation is then mechanical.
> For sure there
> were better attempts than RDF/XML, ones that did not confuse XML
> namespaces (the syntax layer) with naming authorities (the semantic
> layer). The XML Information Set has individual characters represented,
> rather than strings, which makes querying rather more complex than it
> could be.
Existing infoset RDF would work (mostly) with the addition of text nodes 
as shorthand for sequence of characters. I feel that upgrading to an XDM 
RDF model would support more use cases, but I can't say right now what 
those would be.
>
> It might be that a serialization option from XDM instances to RDF could
> be defined.
This is what I am proposing, except to standardize the "what" (schema) 
rather than the "how" (serialization). And the main purpose is to 
support RDF-to-(non-RDF)XML. There are already plenty of options for 
getting RDF from XML.
>
> In general it’s not always easy to see how to go from RDF to XML (for
> example, what to do if RDF property names or values contain non-XML
> characters such as NUL) but i think it's definable if one does not
> require a stable round trip to produce the same XML each time.
Round-tripping is an optional use case; I have never encountered 
it--just as it is rare to want a reversible xslt transformation. In the 
use case I proposed, naming of elements and attributes is controlled by 
the RDF-to-XDM transformation author. If the naming functions and 
structural transformations are reversible, round-tripping would be possible.
>
> So in that case, maybe RDF (JSON-LD?) could be added to serialization.
> How widely do you think it would be used?
I doubt many projects are blocked by lack of a standard XDM RDF schema. 
But a great many could benefit from one--any RDF project with a 
requirement to display formatted views of their knowledge graph. It 
would add a small but important bit of bridging technology that would 
enable greater things in information delivery.
>
> Maybe it would be better done in the expath group?

If this group punts it, maybe so. I see it as closely tied to the XDM 
3.1 standard, which seems in purview of this group.

Regards,
--Paul

>
> best,
>
> liam
>

Received on Tuesday, 16 July 2024 13:52:47 UTC