Tuesday May 06, 2008

Operation SwoRDFish: The Business End

You’ve probably heard from some that structured metadata is what the Web really needs to evolve to the next level. “It’s all about the Semantic Web”. You’ve probably heard from others that structured metadata on the Web will never work because we’re all too lazy, too bad at spelling and too heavily infiltrated by crooks and spies. “It’s all about metacrap”. Who knows how that movie will end, but what I know is that we’ve quietly been putting structured metadata to work on our own little corner of the Web for years. Well, not always so quietly-- you might have heard of Sun’s SwoRDFish project. It’s a pioneering effort to use Web-friendly metadata, and specifically RDF to identify and describe Sun products across systems and departments. I found myself right at the center of things back when the project really gained legs back in 2003 or so, driving the effort to develop a shared taxonomy of products on the SwoRDFish system. That was about the time I became involved in CME and the maturing of Starlight, the main CMS for global sun.com Web sites.

Starlight needed a consistent way to tag pages related to products for a variety of reasons. A customer who follows a link to a Sun success story might learn that the Sun Fire X2200 M2 Server is an effective hardware platform for high-scale Web application architectures, and we want to make sure that they can find on that page well-updated information on that product. On the other hand, if a customer comes straight to a product feature page we want to make sure we make available just the tags for other standard pages associated with that product. We don’t want to have a downloads tab for a hardware product, so we’re concerned about identity and also general categorization of the product. Sun’s product Web pages use SwoRDFish identifiers and taxonomy to solve many such problems, and it’s become a crucial part of our machinery for dynamic page rendering.

For a quick look at the first example in pictures, the following is from our Customer Success Stories landing page:

The “By Product” tab uses selection by SwoRDFish IDs. If you hover your mouse over “Coolthreads” you’ll find the link is something like http://www.sun.com/customers/index.xml?p=8871f410-44cb-11da-ac39-080020a9ed93 . Yes, urn:uuid:8871f410-44cb-11da-ac39-080020a9ed93 is the SwoRDFish URI for the Coolthreads server product category. Clicking on it produces the generated landing page for all customer success stories that touch on that product category.

There is also useful separation of responsibility here. The product metadata and taxonomy is now maintained by experts on the SwoRDFish team, taking their input from product managers. This is general institutional knowledge that is used across Sun, and not just on the product marketing Web pages. Starlight queries SwoRDFish for the metadata, and makes this available for publishers to add capabilities such as “success story by product”. We have built some cool internal tools for content authors to navigate the SwoRDFish ontology to make tagging and publishing quick and simple.

SwoRDFish was definitely on the bleeding edge of things. We learned lots of lessons, and we’d all do things a bit differently if we were starting out now. For one thing, rather than using UUIDs URIs we’d use good old “http://”. We’re actually working on some things in that direction that I hope to be able to talk about more, soon. Perhaps we'll even get to the point of opening up the richness of our data and metadata to publishers, partners, and others outside Sun as well as inside. In this age we realize that our community is our most effective marketing arm. It’s certainly getting easier to carry the small-S semantic web message inside and outside Sun. Now that people see mashups everywhere they understand the importance of connecting the data behind Web sites, as well as linking pages. I definitely see a bright future for the marriage of sound document design and rich metadata design that’s fueled the success of Starlight to date. Sometimes the bleeding edge may cut you, but when it does take root, and proves its value so thoroughly, the satisfaction more than makes up for the scars.


Passionate about data engineering strategy and solutions for Sun’s external web sites. Happiest when building taxonomies, data models, and high performing teams.

Kristen Harris
Web Data Engineering


« February 2017