Curiouser and curiouser!

You'd think with a name like OWL I'd be able to come up with a pun...

Tue, 08 Apr 2003 20:47:00 GMT

Roger Costello and David Jacobs have published a quick introduction to the OWL Web Ontology Language (OWL) which is a semantic layer that sits above the Resource Description Framework (RDF).

[Posted on xml-dev]

I doubt this introduction will make converts of anyone, OWL just isn't that easy to grok, but the applications it presents are credible ones. I think the problem for skeptics is that they (a) have seen it all before, and, (b) are suspicious of things that they can't directly grapple with, and, OWL & RDF are not for the fainthearted.

The fact that RDF is not widely considered to be a success is not surprising to me. It looks complicated and it's not immediately obvious what you do with it - it's just metadata after all. I believe that OWL or, more precisely, useful OWL ontologies, are what will make RDF successful. It's OWL that tells applications what they can (and cannot) do with RDF metadata. That's it's power.

Two key things need to happen now:

The creation of applications that help developers to create robust and useful ontologies.
The creation of libraries for embedding RDF+OWL into applications (as ubiqitous as XML parser libraries are today).

If these two pieces of the puzzle can be assembled then I think we will see a raft of components built that utilize ontologies to perform their functions. These components will be made into JavaBeans, .NET assemblies, published as web services, whatever. Users will consume those services and never need to know that OWL or RDF are involved.

However until those tools arrive I believe the only course left to you (unless you're trying to build them yourself) is to be pragmatic and keep a weather eye open. As Dave Winer commented recently about RDF: "I am not against it, as long as it doesn't interfere with what I want to do."

Giants go 7-0

Tue, 08 Apr 2003 10:28:59 GMT

Giants go 7-0! Woo Hoo

Franchise history is being made, now if we can just make it to 10-0!

Insight from David Gurteen.

Tue, 08 Apr 2003 09:39:03 GMT

Results Oriented KM. Some recent thoughts on KM:

KM is not an end in itself. It is a set of disciplines and tools to help us meet our business objectives.
What is the point of doing KM if it does not help us meet our business objectives? KM can only be measured by its ability to help us meet our business objectives.
KM needs to address the quality of our decision-making.
What is the point of KM if we still make lousy decisions - if we do the wrong thing - even exceptionally well? We would do better to do the right thing badly and not bother with KM at all!
KM needs to address the issue of our motivation and our ability to make use of the knowledge we have.
We can be given all the perfect information and knowledge that we need to do our jobs but if we fail to use it then what is the point?
KM should help us to improve our awareness and understanding.
KM should not be just about helping us to know more. It is through increased awareness and understanding that we start to see our organizational world in new ways and identify new business opportunities.
KM is about helping us to identify new opportunities and leveraging them.
Measuring the results of KM is important but we should not forget that KM is also about identifying new opportunities. We can measure cost/profit etc but we cannot measure 'missed opportunities" by their very nature we do not see missed opportunities until it is too late. [Gurteen Knowledge-Log]

Some insightful thoughts from David.

liveTopics go fast now

Mon, 07 Apr 2003 22:27:26 GMT

I'm actually pretty pleased with this.

I wrote a new Table of Contents driver for liveTopics which uses the new OPML generator.

Now, fully regenerating my table of contents (the whole 239 topics A..Z of it) takes 49 seconds start to finish!

This was a task that took well in excess of 16 minutes (not counting upstreaming) last time I tried it.

Pepping up liveTopics performance

Mon, 07 Apr 2003 19:40:01 GMT

I knew that Radio's outline were slow but...

Internally liveTopics uses Radio's outlines quite a lot. When you are outputting outlines for activeRenderer it seems quite convenient to use Radio's built-in support them as a type. But as my topics database has grown, liveTopics has begun to groan. And this on a pretty zippy P4 1.6GHz, I can only imagine what it must be like for some people out there with somewhat slower machines (Marc?)

So I took a first cut a few minutes ago at re-writing the liveTopics outline generator and did a little benchmarking. At the moment it takes the Radio outline based generator 64 seconds to generate the XML for my topic liveTopics which is the largest topic in my database at 82 posts. By contrast it takes my hand-rolled OPML generator just 2 seconds to do the same work!

It will take a little bit of time to replace the current generator properly but I anticipate the performance difference will be startling.

Topic Rolls, Controlled vocabulary and the ITE

Mon, 07 Apr 2003 16:51:49 GMT

Topic Rolls near reality.

The next step is to allow liveTopics to import topicrolls from other locations. [Matt Mower]

Matt expands his work on topic information for Weblog entries. He adds another hierarchical layer by introducing "topic types". This basically allows you to group your containers (topics) into another layer of containers (topic types), thus adding more modelling power to one's own topic structure. Without knowing what really is going on in Italy I would be very cautious with attempts to "define a control language for topics rather than allowing users to make up their own." This is where most projects in this are seem to go the wrong way. At some point someone steps in and wants to take control instead of developing tools that would allow the negotiation of shared topic meanings... [Sebastian Fiedler]

[Seblogging News]

Since Paolo's currently out of range I guess I should add something here.

Blog Notes (Giuseppe Granieri's blog aggegrator project) is a structured news feed viewer. It allows you to attach certain pre-defined topics to a post to indicate such things as:

the post is an ironic comment
the post is about politics

In order to keep the structure under control the range of topics must necessarily be restricted. So in essence the range of topics in Blog Notes forms a controlled vocabulary and liveTopics will, to some extent, support this.

However that won't mean that people will be unable to create their own topics or interact with other applications that do permit is such as Phil Pearson's Internet Topic Exchange. As a sidenote I am very conscious that the work in integrating liveTopics and TopicExchange never got finished. I have it in mind to get in touch with Phil again and see where we are with the outstanding issues. (Phil if you see this and have a moment, please ping me via IM).

Is RSS1.0 Taxonomy module dead?

Mon, 07 Apr 2003 16:22:46 GMT

I'm really interested in finding out whether anyone actually uses (either produces or consumes) the the RSS1.0 Taxonomy module (hereafter <taxo:topics>). If you do, I would be grateful if you could add a comment to this post along with the URL of your feed and/or application.

As I posted on Saturday I am investigating methods for incorporating topic metadata into RSS feeds. As a Radio developer I'm mainly interested in RSS2.0 feeds but I don't want to duplicate any efforts / re-invent any wheels unnecessarily. So I'm considering <taxo:topics> very carefully.

However my view is that almost nobody is using it. I've spent a good bit of time this afternoon examining the feeds of various people in the RSS community (obviously trying to home in on those using RSS1.0) and there is no sign.

Phil Ringnalda points me at Syndic8's feed information where you can see that even while 25% of feeds there are in the 1.0 RDF format, less than 2.5% actually declare the <taxo:topics> namespace (and we have no way yet to verify whether they actually go ahead and use it).

If it really is the case that nobody is using it then I have to ask why?

AntiCap UK

Mon, 07 Apr 2003 16:01:39 GMT

AntiCap lobby group calls for BB freedom. 'What don't we want? Caps. When don't we want them? Now' [The Register]

A good site to keep an eye on for anyone using broadband in the UK. I am a Telewest cable customer and have been very happy with their service so far. But if they futz with the T's & C's...

When Google news isn't

Mon, 07 Apr 2003 15:48:04 GMT

'Adios Google'. Letters Readers repelled by PR subterfuge [The Register]

It appears that Google have decided that company press releases and lobbyist puff pieces are legitimate news items as far as Google news is concerned.

I didn't like that movie either...

Mon, 07 Apr 2003 15:18:58 GMT

Trooper-style citizenship – as Heinlein satirizes – is a condition that sounds free and honorable but is actually impossible to exercise with free will or honor. A recent article on posthumous citizenship for some non-American soldiers speaks volumes when it notes that the citizenship is not real or practical, but symbolic. It suggests citizenship may be most meaningful to the dead, the static, the non-thinking.

Heinlein’s citizenship is granted for soldiers who have made it through boot camp, where they have learned not to question authority, to follow all orders from above instantly and exactly, and who have no other allegiance than to the all-wise central state. It is a Rumsfeldian vision of citizenship. It is a citizenship where each moral compass is not individually discovered, tested and mapped, but instead simply imprinted. It must be because "Man has no moral instinct."

[From LewRockwell.com]

Recently retired USAF lieutenant colonel Karen Kwiatkowski compares the Bush/Rumsfeld doctrine to Robert Heinlein's vision in Starship Troopers and finds some uncomfortable similarities.

Exploring topics in RSS2.0

Sat, 05 Apr 2003 10:38:33 GMT

I've been doing some thinking about how to encode topic information into RSS2.0 feeds. As a simple test of the Radio callback facility I have implemented a very simplistic protocol. Within each <item> is a tag

<topic id="topic_id" type="topic-type" source="url">topic name</topic>

for each topic associated with the item (post). A concrete example (using the rsstopics namespace):

<rsstopics:topic rsstopics:id="the_state" rsstopics:source="http://matt.blogs.it/topics/topicsT.html#the_state" rsstopics:type="generic">the state</rsstopics:topic>

Whilst this does have the advantage that it's simple and direct it's also a bit silly to invent a new format for topic information when we have two standard culprits available already:

RDF is a general format for describing resources. A resource in RDF terms is anything which can be uniquely identified by a URI. An RDF statement (utilizing Dublin Core metadata) that asserts me as the owner of my weblog might look something like:

<rdf:Description rdf:about="http://matt.blogs.it">
<dc:Creator>Matt Mower</dc:Creator>
</rdf:Description>

If you cut away the syntactic fluff what this says is:

Matt Mower is the Creator of http://matt.blogs.it

Referring back to the problem at hand, describing what a post (expressed as an RSS item) is about we could come up with something like:

<item rdf:about="permalink">
<topic id="topic_id" type="topic-type" source="url">topic name</topic>
</item>

Which is more or less exactly where we started -- using RDF hasn't altered the solution but it has added some framework around it (in this case adding rdf:about to signal the presence of RDF data within the item). However we can go a step further. A useful article by Eric van der Vlist discusses this very subject and refers to the RSS1.0 taxonomy module.

Somewhat counter to what you would expect RSS2.0 does not follow on from RSS1.0, nor does RSS1.0 follow on from the popular RSS0.9x formats. RSS1.0 is, depending upon your point of view, a step forward or an aberation. RSS1.0 uses a modular set of RDF based tags to describe items in the RSS feed. One such module is the Taxonomy module which is intended to allow classification of RSS channels & items.

Using the taxonomy module you create something like:

<item rdf:about="permalink">
    <taxo:topics>
        <rdf:Bag>
            <rdf:li resource="topic-uri-1"/>
            <rdf:li resource="topic-uri-2"/>
        </rdf:Bag>
    </taxo:topics>
</item>

Here the <topics> element contains a list (using the RDF defined Bag - or unorderer list - container element) of resources indicating topics that describe the item. Each resource then has a <topic> element that describes the topic. It might look something like:

<taxo:topic rdf:about="http://matt.blogs.it/topics/topicsT.html#the_state">
    <taxo:link>http://matt.blogs.it/topics/topicsT.html#the_state<taxo:link>
    <rsstopics:type>generic</rsstopics:type>
    <dc:title>The State</dc:title>
</taxo:topic>

Although it's a jumble of RDF, the RSS1.0 taxonomy module, Dublic Core, and, a custom rsstopics schema this says exactly the same thing as the original:

<topic id="topic_id" type="topic-type" source="url">topic name</topic>

But do we have to deal with such an ugly mess? Perhaps not. Our original choices included the XML Topic Maps format. This is a complete specification for exchanging topic information. An example of a topic in XTM format might look something like:

<topic id="the_state">
    <instanceOf>
        <topicRef xlink:href="http://www.purl.org/rss-topics/rss-topics#generic">
    </instanceOf>
    <baseName>
        <baseNameString>The State</baseNameString>
    <occurence id="the-state-item">
        <instanceOf>
            <topicRef xlink:href="http://www.purl.org/rss-topics/rss-topics#story">
        </instanceOf>
        <resourceRef xlink:href="<permalink-uri>">
    </occurence>
</topic>

Again this encodes the same information, using a standard format and only one required namespace (that of XTM itself). A URI such as http://www.purl.org/rss-topics/rss-topics#generic points at a topic in another map (in this case a topic describing the topic-type generic).

The use of XTM comes with a number of advantages with the main one being that there are an increasing number of tools available to process & manipulate it (for example, see topicmap.com). However there also a number of problems with this representation when you attempt to embed it within another XML format such as RSS.

It's not clear whether an XTM fragment such as this is valid when used in this way
Each time a topic is used we will be duplicating it's details, bloating the markup & potentially creating invalid entries
The <occurence> relation within the <topic> element is technically redundant. The enclosing <item> indicates the occurrence.

One way to avoid these problems would be to embed the topics within the RSS <channel> definition and refer to them from each <item>. However we still need a way to refer to the topic and XTM doesn't provide this. If we had a good way to reference topics then we could either embed mini topic map within the RSS file, or just have the <topicmap> in an external file and point to it. What could we use? One possibility is RDF.

Using a combination of RDF and XTM would mean something like:

<item rdf:about="<permalink-uri>">
<rsstopics:topic>http://www.example.org/myTopicMap.xtm#topic-id</rsstopics:topic> 
</item>

<item rdf:about="<permalink-uri>">
<rsstopics:topic>#topic-id</rsstopics:topic> 
</item>

In this example the item now refers to an XTM defined topic either elsewhere in the RSS feed (contained within a valid <topicmap> element) or within an external topic map. The referenced <topic> element can further describe the topic (names, types and so on) using all the expressiveness of XTM. It's also efficient since there is no duplicated information within the feed.

I have described approaches using RDF, XTM and a hybrid of the two. Each has advantages and disadvantages although I believe the hybrid makes the best use of both formats.

I'd welcome comments and or opinions from interested parties.

The slippery slope

Thu, 03 Apr 2003 22:57:29 GMT

Intel consultant detained without being accused of a crime. No right to a lawyer. No chance to see the evidence against him. Secret proceedings. Covered here. What country do... [BookBlog]

Slippery slopes are... well... slippery.