N-Triples is plain-text serialization of RDF, nothing more. This is RDF's problem. 

JSON's RFC (4627) says "JSON text SHALL be encoded in Unicode. The default encoding is UTF-8."

Again, why wouldn't you be using a JSON parser/serializer to handle this? And if it's RDF/JSON (or JSON-LD), again, it's RDF's problem. 

"RDF in HTML guise" (I assume this is referring to RDFa?) would defer to the charset declaration of the page, as would any other HTML-based serialization. 

HTML is certainly likely to be the most error-prone (since it's also the most democratic), but it, again, is an "HTML problem". If the character encoding matches the declared charset and it's valid HTML... what else can you do?


On Monday, January 7, 2013, Andrew Cunningham wrote:
if RDF is going to be used exclusively, but if you have N-triples and JSON in the mix as container formats ... different issue.

And no, they aren't RDF's problems, RDF in XML format inherits a lot of features from XML and can leverage off ITS etc. Likewise RDF in HTML5 guise can leverage off internationalisation features in HTML5 or HTMLNext (Living HTML or whatever its called this week). The issues are more related to bibframe and the conversion process from MARC formats to Bibframe regardless of the container.

For instance RDF and XML would use one system for language tagging a record, and MARC and possibly. Bibframe use a different system for tagging the language of the item/object being described.

Two different functions and two different language tag schemes ... BCP47 vs ISO-639-2 (B)

And in theory the record might consist of multiple "language" tags since in script and romanisation would be different language tags in the BCP-47 sense.

This level of complexity would become an issue when records are being transformed into XML or HTML5 formats to be used by user agents, since accessibility requirements will kick in in various jurisdictions.

It will also impact on font rendering of content, in IE10 and latest versions of Firefox content marked up with a language tag of "tr" will kick in The Turkish language system in the font if present in the fonts OT tables, etc.

At least that's my high level tack on it.

In MARC-8 and MARC-21 you didn't have to concern yourselves with this, they essentially lived in isolation. In theory internationalisation was based on a 40+ year old model.

But RDF in XML, RDF in HTML5, N-triple and JSON each bring their own requirements to Bibframe;

As do the programming languages used;

Accessibility requirements;


Bibframe is movement into a model where there are many inter-dependencies and external requirements on the model. Going from an isolated industry standard to leveraging off international standrads


On 8 January 2013 13:36, Ross Singer <[log in to unmask]> wrote:
Why are they issues, though?  They're RDF's problems, not Bibframe's. Isn't that part of the point of using existing standards?


On Monday, January 7, 2013, Andrew Cunningham wrote:
Although those legacy encodings specific to the library industry would not exist in Bibframe

Ultimately the issues are more related to how the parsed content is going to be consumed or going to be used. If it is to be human editable or presented to user agents then more complex processing that inserts markup or formatting control characters that are not present in the MARC records would sometimes be required.

A lot of this is just tip of the iceberg ... esp if transforming to HTML5, language tagging in the record versus language tagging of the record, and a range of other issues.


On 8 January 2013 12:36, Ross Singer <[log in to unmask]> wrote:
On Monday, January 7, 2013, Andrew Cunningham wrote:
And it got me to thinking, a point quoted in the article was that each of the implementations is doing different transformations on the MARC records

The second point is that RDF/XML, N-triples and JSON formats are supported.

One markup format, one plain text format, one javascript format.

Which got me to thinking, each of these formats has different requirements; for instance:

* RDF/XML would use markup where N-triples and JSON would use Unicode Formatting Control Characters.

* RDF/XML and N-triples would reference characters outside the Basic Multilingual Plane directly as characters or as six digit hexadecimal numerical entities, while JSON requires to four digit hexadecimal numerical entities representing UTF-16 surrogate pairs.

* RDF/XML can use characters directly or XML/HTML style hexadecimal or decimal numerical character references or named entities (e.g. &#x0100;) while JSON requires javascript nuerical entities ,e.g. \u0100; finally N-triples is more agnostic but has some interesting requirements, e.g. requires support for all Unicode characters and references charmod, and indicates a preference for actual characters over escaped characters, except where required by the encoding.

So different intermediation processing of characters maybe required for each format, as well as logic to handle markup versus Unicode format control characters.

If this makes sense?

It does, but I'm not sure why it matters?  It's all RDF and presumably one would be using RDF parsers to handle the character encodings. 

I mean, we deal with this already with MARC8, UTF-8 (in MARC-21), and MARCXML. It's only really a problem because we use encodings that nobody else in the world uses so we have to come up with our own parsers and serializers (and, in many languages, MARC-8 support is just ignored). 

The RDF community is already dealing with this (plus other serializations), so I don't really see how this is an issue. 

Although, admittedly, I may be missing your point here. 


On 8 January 2013 08:55, Andrew Cunningham <[log in to unmask]> wrote:

I hate to tell you this but numbers aren't language neutral.

But there are bigger internationalisation issues and poten

State Library of Victoria
328 Swanston Street
Melbourne VIC 3000

Ph: +61-3-8664-7430
Mobile: 0459 806 589
Email: [log in to unmask]');" target="_blank">[log in to unmask]