It’s just data

RDF/XML Working Draft feedback

On Thursday, I talked with Dave Beckett on #echo about possible changes to the RDF/XML Syntax Specification.  The current status of RDF/XML is that it is in a W3C Last Call Working Draft form - meaning that only minor corrections are accepted at this time.  My inputs clearly go beyond that, but I will share them anyway on the off chance that they resonate with others and/or can be taken as input for the next round of the specification.  I've ordered them roughly in order of priority:


I don't think this is going to work. These features are all very small transformations, and being so restricted will make the resulting rdf sub-optimal.

An example from Atom: the author element can be added to each entry. But as because the author is often the same for each entry, Atom has a shortcut that allows you to add the author once to the feed element. You wouldn't want such shortcuts in your RDF model.

I've updated the XR transformation to reflect this. (http://w3future.com/tools/atom2rdf.xml) An author element in the feed is transformed to a foaf:maker property on each entry, unless this entry has its own author element.

Posted by Sjoerd Visscher at

My hope is that the RDF WG will finish work on this set of drafts and call them done so that there can be some measure of stability.

Once stability gives people enough sense of confidence, I hope they start to experiment with other approaches for serializing RDF and extracting RDF from other formats. RDF/XML has a much smaller number of fans than RDF, in my experience, so I think people will be interested in doing this.

I suspect some of those experiments will address the concerns you have above, and maybe someday percolate back into the mainstream of RDF work.

Posted by Simon St.Laurent at

Since the RDF/XML effort has been ongoing for over two years, and we're finally at last call, I also doubt that there will any interest in making further modifications at this time. Each entry has been discussed numerous times and in numerous conversations. Every element is there for a specific reason, and after much debate. Changing it so taht it's more 'palatable' for the Pie/Echo/Atom folks, or XML purists isn't going to cut it with those of us who've wanted a stable specification.

I disagree with Simon on RDF/XML having a small base and eventually fading away. The other alternatives that float around are primarily used by the R & D'ers, the W3C knew that they needed to use a serialization format that was at least somewhat familiar, hence RDF/XML, rather than so esoteric, lisp-like formatting like N3.

It really is only the XML purists and the prejudiced that will continue to resist. The rest of us are adaptable.

Posted by Shelley at

Hmmm, last sentence came off borg-like. Replace with:

"It is only the XML purists and prejudiced who will continue to pick at the XML serialization format for RDF. The rest of us will see the potential, overlook any of the flaws, and make it work, because we're adaptable."

Better.

Posted by Shelley at

I think you will get pushback on any suggestion at this point that isn't a bugfix. The wg has always been strongly reluctant to perform surgery on the syntax, even back when I was a member.

I agree with Simon. All things considered, Dave Beckett and the wg have earned the recc. I would let this ship and start over.

Posted by Bill de hÓra at

Substitute "people who care what the XML looks like because they actually work in it" for " XML purists and the prejudiced", and I can agree with Shelley.

I think there are a lot more of the former than the latter, though.

Posted by Simon St.Laurent at

"It is only the XML purists and prejudiced... "

Purists and prejudiced, eh? I don't really need to quote more than that, do I?

Posted by Joe at

Regarding rdf:parseType - Not sure how you'd reduce or eliminate this - its only used in very specific non-default circumstances, and would probably only be seen in PEAW as rdf:parseType="Literal" and rdf:parseType="Resource", with good reason in each place.

Regarding unprefixed attributes - Would a default namespace not be suitable here? RDF/XML is just following XMLs lead here, and unprefixed attributes do not inherit their parent element's namespace in XML. I doubt RDF/XML should diverge from this behaviour in any way. It is an XML syntax, after all.

Regarding a parseType for URIs - the attribute is called rdf:resource. :) In certain situations, rdf:about may also be of value.

Posted by Jeremy Gray at

RE: RDF/XML Working Draft feedback

Jeremy,

Regarding unprefixed attributes - Would a default namespace not be suitable here?

Unprefixed attributes do not pick up default namespace declarations (unlike unprefixed elements which do) so I don't see how this would help.

Message from Dare Obasanjo at

Jeremy: the declaration of a default namespace does not affect how an attribute is to be interpreted.  In other words, it is always the case that an attribute named 'attr' and an attribute named 'ns:attr' are distinct independent of the value of the default namespace and the current association for the 'ns' namespace.

However, XML is just a syntax.  It is quite OK for an individual grammar to declare that certain combinations of attribute and elements are valid, and certain other combinations are invalid.  An example would be the name and http-equiv attributes in the meta tag in HTML 4.0.1.

A more general comment to Jeremy and others: there is plenty of room for more proposals.  If you believe that you have a more suitable syntax to propose that is either more metadata rich or simpler, please do so.  If, for some reason, it isn't practical for you to do so on your site, feel free to drop the proposal in the wiki.

Posted by Sam Ruby at

RDF, XML and Interoperability

The Pie/Echo/Atom/Whatever discussion is bringing up the RDF/XML debate... yet again. On one side are the folks that say PEAW should use RDF/XML serialization, on the other side are the folks that say PEAW should just use some vanilla XML...... [more]

Trackback from snellspace

at

Yes, Joe -- purists and prejudiced -- with clarification on prejudiced being those who will see no good out of RDF/XML regardless of demonstrations, writings, tutorials, possibilities, potential whatever. By both of these I mean people who will believe what they will about RDF/XML no matter we do, no matter what benefit we demonstrate. 

Words I stand by. But if you want another definition of same see these.

Posted by Shelley at

RDF Latest News

Sam Ruby has a moan about RDF / XML. Some people just don't understand. Also Edd Dumbill has a little gripe with RDFDrive, the only compliant RDF parser available for the .Net platform so I suppose beggars can't be choosers. 'Why all the...

Excerpt from Web Service Guy at

Shelley,
  And if you read those articles I've written you'd see a clear practice of seperating my criticism for the XML serialization of RDF from any comment on RDF itself. I been consistent in my criticism of the XML serialization of RDF and if being consistent is the same as being prejudiced in your eyes then so be it.

  I have also consistently raised concerns that the RDF advocates hold forth that every format must be RDF/XML or nothing, this latest foray into Atom/RDF is just another  example, even with a working XSLT transformation from Atom to RDF there are claims that this is insufficient. If repeatedly raising that concern gets me the label of being a purist, so be it.

Posted by Joe at

Shelly, I'll be the first to admit that I am prejudiced against RDF syntaxes and most RDF initiatives.

So what?  I am also prejudiced against snake oils.

My prejudice is founded on apparent problems with RDF syntaxes and past RDF initiatives and discussions.

I won't mind all that if the supposed 'enlightened' RDF proponents stayed within their caves instead of trying to spray on a coat of RDF on every XML format they see, effectively forcing me to use RDF.

Theoretical benefits of RDF are unnecessary to Atom.  From that perspective, I see no difference between RDF and snake oil.

Posted by Don Park at

But it wasn't the RDF/XML people that brought up the possibility of Atom being RDF this week. Aaron, but Aaron's never liked RDF/XML, himself. I would rather the possibility of Piet/Echo/Atom had not been brought up at all then open the door for yet another round of 'bad RDF/XML', 'evil RDF/XML'.

Pie/Echo/Atom syndication feed go on its merry little way to an isolated XML vocabulary. I'll stick with RSS 1.0.

Posted by Shelley at

Shelly, I don't fault Aaron for wondering out loud.  People do that and I think it is good.  Heck, even I revisit RDF occasionally when the fancy hits me.

I also don't fault you nor other RDF folks for doing what you believe in just as I am doing what I feel I must do so I can be more lazy in the future.

Shape of the future is tempered with fire and pounded into shape between anvil and hammer.

Posted by Don Park at

I have to believe that at one point, Dave Becket looked at the 1999 RDF Model and Syntax Specification and thought to himself, "I bet I can make this better".

Posted by Sam Ruby at

In two years someone will be looking at Pie/Echo/Atom and saying, "I can do this better". A couple of months ago a group of people looked at RSS 2.0 and said, "I can do this better". No matter what anyone comes up with, there will always be people saying, "I can do better".

A decision was made in 1999 to continue with RDF/XML, even though it was putting a graph model on a tree structure. The reason why? Lot's of talk about XML being the format of the future within the W3C and so on, but let's face it -- if N3 had become the accepted format, you wouldn't have even a tenth of the interest in it now. It would be too different, too scientific looking -- too esoteric. Scientifically it might be better. From a marketing stand point? It's a bust. It's geek talk. RDF/XML, for better or worse is something that normal people can get their mind around -- not necessarily the syntax, few people need to work directly with; but the fact that it's XML. They 'know' XML.

It is that common touch of familiarity that was essential for RDF. Without it, it would be the Eiffel of the Semantic Web. The QUEL of databases.

I like RDF/XML. It's more precise than vanilla XML. As we found out -- now are contributors a list, or a unordered group? I never went into RDF/XML with the baggage of XML and all it means behind me. No sacrifices on the alter of the railroad crossing. Because of this, I've had no problems working with RDF/XML. Not a bit. Same with others. Regardless, though, know this -- RDF/XML is not going away. And the RDF working group is not going to throw out RDF/XML, or alter it significantly now, because the Pie/Echo/Atom folks don't like it.

A sense of perspective is, I think, way overdue now.

Way overdue.

Don't like RDF/XML? Fine. Don't use it. But don't deny me my right to use it, to want a stable specification release, and to support it -- including supporting a syndication format that is based on RDF/XML, RSS 1.0.

Good luck with Pie/Echo/Atom, Sam. I'm sure there are RDF/XML people who still have some patience with the game and will help you develop lovely XSLT transforms that are just so easy to run.

Posted by Shelley at

Taking something which is currently illegal (e.g., attributes without a prefix - my second example) and providing a meaningful interpretation to this is backwards compatible.  My only hope in presenting this list was to get them considered.  I discussed this with Dave Beckett.  His response wasn't "these are bad ideas", but something more along the lines of "not this release".

It may turn out that one or more of these are bad ideas.  But IMHO, they shouldn't be rejected out of hand.  And perhaps the underlying requirements that motivated these ideas can be ferreted out and a better solution can be found.

Posted by Sam Ruby at

Re. RDF/XML in general - it may be too ugly to appeal to many developers used to simple XML (non-namespaced, possibly full of treble-escaped stuff). But RDF/XML works, and doesn't take that much getting used to.

As Simon suggests, at this point in time I think the stability is the priority over any aesthetic concerns.

re. the three specific suggestions - the area around
rdf:parseType might be tweakable, but the reasons for changing it are IMHO not very strong: if document size is that much of a problem, then use a different serialization (or zip); if readability is an issue, errm, then use another serialization... RDF is aimed primarily for machine-readability, making it pretty isn't that important.

Namespaced attributes - I believe this is a failing of XML namespaces rather than anything specific to RDF.

"Introducing an rdf:parseType that indicates that a given attribute or element represents a URI..." Within RDF/XML the roles of attributes and elements aren't arbitrary, and there already is the (vital) mechanism for referring to URIs. But I suppose it would be possible to type the content of a Literal as an xsd:anyURI, but then it would open up a can of worms with the distinction between resource and literal nodes in the graph.

It's pretty apparent that all three of these suggestions came from trying to mould current Atom XML into RDF, which is perfectly reasonable, and I don't think this moulding task is in any way an edge case for RDF/XML. But I do think the syndication domain does bring with it fringe quasi-requirements, I can't think of any other domain where such a fuss would be made about the human-readability for example. I doubt whether saying this will change any attitudes, but these things derive more from expectations based on previous formats, rather than any genuine need derived from business analysis.

Posted by Danny at

Danny is correct about un-prefixed attributes being an XML Namespaces problem, not an RDF/XML problem. The RDF WG can't solve it, I don't think.

I quite remember how appalled I was when I had it explained to me that no, an unprefixed attribute is not assumed to be in the same namespace as its element.

IMO it's a grotesque wart on namespaces. Sam, I'm honestly not sure whether proclaiming that unprefixed attrs are to be treated as if they belonged to X namespace (for any value of X) is legal, but what the heck, give it a try -- if it isn't, you'll find out soon enough.

Posted by Dorothea Salo at

Dorothea, while I would not disagree that this could be perceived as a wart, it certainly is legal for any specific grammar to specify a validity constraint where certain attribute combinations are mutually exclusive.

To give a concrete example, the following well formed XML could be declared invalid according to the rdf schema:

<rdf:Description about="http://foo.org" rdf:about="http://bar.org"/>

This turns out not to be a hypothetical example - see how the current working draft describes how to handle the Attribute Event.

Posted by Sam Ruby at

It's possible to add constraints to validity for XML languages, but not remove them and stay within the confines of the underlying specs.

Treating unprefixed attributes as if they were in the namespace of their parent elements is not consistent with XML Namespaces. I agree it should have been done this way, but RDF or any other syntax can't be interpreted in that way and still call itself XML.

See Namespace Myth #4 :
http://www.xml.com/pub/a/2000/03/08/namespaces/index.html?page=2

Posted by Danny at

Danny writes:
"Treating unprefixed attributes as if they were in the namespace of their parent elements is not consistent with XML Namespaces. I agree it should have been done this way, but RDF or any other syntax can't be interpreted in that way and still call itself XML."

Maybe. RDF is processing the XML anyway. If RDF wanted to, it could specify processing like:

http://simonstl.com/projects/namespace/

using a mechanism like com.simonstl.namespace. attributes.AttributeNamespaceFilter.  That doesn't affect the XML nature of the original document, but does specify that RDF is going to read it a particular way.

Heck, it's not like RDF cares about entity references or quoting styles in the original XML document, so I have a hard time seeing this as that much greater a reach.

Posted by Simon St.Laurent at

Blindsiding and Forward Thinking

When Sam Ruby invited me into an IRC chat last week to talk about creating a compliant RDF/XML syntax for Pie/Echo/Atom, I wasn't aware that there was an agenda behind this effort. Not until Sam published the full model at his weblog and discussed how... [more]

Trackback from Burningbird

at

RDF Latest News

Sam Ruby has a moan about RDF / XML. Some people just don't understand. Also Edd Dumbill has a little gripe with RDFDrive, the only compliant RDF parser available for the .Net platform so I suppose beggars can't be choosers. 'Why all the...

Excerpt from Web Service Guy at

Add your comment