LC Response to comments from Ian Hickson for RDFa Core 1.1 - ISSUE-66

Hi Ian,

This is a response from the RDFa Working Group to your last call 
comments filed under ISSUE-66.

Responses in-line from here.

RDFa Working Group Issue Tracker wrote:
> RDFa Core 1.1 Last Call comments from Ian Hickson:
> 
> I believe RDFa should not have prefix-based indirection, for reasons that 
> have been listed many times over the past few years; since this draft adds 
> two new indirection mechanisms rather than removing any, I am not 
> satisfied that my earlier feedback has been adequately addressed.
> 
> The permitted use of xmlns="" as one of the prefix declaration mechanisms 
> specifically will cause serious compatibility issues for reasons that have 
> similiarly been elucidated numerous times over the past few years.

As a group we have discussed this many times, both internally and with 
members of the broader RDFa, Semantic Web and HTML communities. So far 
we have had a fairly unanimous response from those that use RDFa, 
telling us that they need some form(s) of URI shortening - aka 
prefix-based indirection - in RDFa.

Due to concerns around the use of xmlns as one of the prefix declaration 
methods, we have now deprecated the use of xmlns, keeping only a minor 
reference to it for processors in step 4 of section 7.5, in order to 
ensure backwards compatibility with deployed RDFa.

In order to address the needs of authors - such as allowing the use of 
properties identified by URIs and defined in various vocabularies around 
the web, and referencing them in a shortened, human friendly form - RDFa 
core introduces the @prefix attribute, which along with CURIEs allows 
URI shortening of the form "foo:bar".

Additionally, it was found that many authors would like to use simple 
"terms" to reference properties from various vocabularies. To cater for 
this RDFa Core introduces the @vocab feature which allows the use of 
simple string tokens - of the style "bar" - in various RDFa attributes.

Prefix based indirection, is a feature which is both heavily deployed 
and found to be useful by authors. RDFa Core 1.1 gives users the choice 
between using full URIs, simple terms, or prefix based indirection, time 
and real world usage will show us all what the people find to be most 
useful.

> The language as a whole is overly complicated, e.g. with a preponderance 
> of attributes significantly beyond the minimum necessary. Again, this 
> issue has been described in detail in the past.

RDFa serves to allow the expression of metadata in attributes, 
specifically RDF in attributes. RDFa therefore defines attributes which 
cater for the full expressiveness of RDF (such as @datatype and 
@content), others which cater for core author requirements (such as 
@prefix and @vocab), and finally it adopts some attributes which are 
well defined and used in many existing specifications (such as @href and 
@src).

In total, RDFa includes 13 attributes, 6 of which are key to the 
expression of RDF, 4 of which are heavily used attributes adopted from 
other specifications, and 3 of which pertain to the use of URIs as 
identifiers, a core feature of both RDF and the Web in general. As a 
Working Group we have carefully analysed each attribute defined in RDFa, 
and found each one to be needed in order to allow the reasonable 
expression of RDF in attributes.

> IMHO, you should do some real usability studies (one-way glass and 
> everything) like we did with microdata. Or at a minimum, see if people 
> outside the RDF/RDFa community can actually use RDFa for any non-trivial 
> purpose without help and actually get it right. Or indeed, the same with 
> implementators. So far, every implementor I'm aware of has done it with 
> fundamental bugs in their first attempt. Some have since fixed those bugs, 
> but any technology can be implemented correctly _eventually_ -- whether 
> the bugs get shaken out after years of work isn't the interesting 
> question. What matters if you want long-term interop is whether someone 
> you've never heard of can implement the technology more or less correctly 
> the first time, without being pestered into fixing problems.

We agree, and have seen wide spread real world usage of RDFa, and 
consequently received feedback from numerous implementers. RDFa Core 1.1 
is the product of that feedback, invaluable feedback from users on scale 
far larger than we could hope to achieve from any preliminary usability 
studies.

RDFa is today being used by various parties, ranging from single users 
applying licenses to works and including metadata snippets in their 
documents, through to large corporations using RDFa for a wide range of 
purposes. Indeed entire industries, such as the News sector, are now 
moving to adopt RDFa.

That is to say, RDFa is not an experimental metadata technology where 
fundamental design decisions can be changed, rather, RDFa is a deployed 
W3C Recommendation which builds on the core design decisions 
incorporated throughout the Semantic Web technologies.

Interoperability is of the utmost importance to us, and as a group, we 
hope that the standardization process, including feedback from 
implementers and specification reviewers, will serve to provide a 
recommendation which can be implemented correctly, the first time.

Many thanks for your feedback Ian, and we look forward to working with 
you in the future when standardizing HTML+RDFa1.1.

Best Regards,

Nathan Rixham, on behalf of the RDFa Working Group.

Received on Thursday, 10 March 2011 21:31:40 UTC