bestkungfu weblog

Whither XML (original copy had the typo “Wither XML”)

The dream is universal access to information. Doing better between people, but not between processes. In 1996, there were two dreams one was documents for everyone (separation of information from presentation). “It was a way to take back” from closed formats “that broke the moment you hiccuped.”

Self-describing data (e.g., elements and schemata in XML) are magic. You couldn’t swap things around in relational databases. You could only work around it. The primary-key-foreign-key model was really just a way to traverse links.

Google is the way we navigate the Web. Faster even than the built-in search mechanisms of sites like Amazon.

It is the best of times:

• XML is self-describing
• Metadata doesn’t require custom parsers
• Amazon and eBay can do $100s of millions of business today with it
• Documents, not rows. Relational databases were designed to serve rows of columns, not documents. Documents aren’t flat; they’re complex. One can now ask for portions of documents
• Web Services are open standards

It is the worst of times:

• ASCII, not information. We’re only serving bits
• Complexity, not KISS: Web Services, Schema, mapping, WSDL, WS-CO, WS-TX, WS-Policy: “I am amazed by the complexity, the difficulty.”
• APIs from Hell: Confusing the message with the messenger. Too much abstraction from a purchase order to a WS implementation
• Coarse-grained, but not contractual: Humans are imperfect, and have to make things work. We can’t describe contracts, or say “call this web site before you call this one”, in a language
• Synchronized, but message-based: This leads to fragility. Synchronization requires low latency. Better to be able to push out data when thresholds are reached than to get millions of people pinging every 10 seconds.
• Methods, not queries: APIs are not as flexible as queries are by design. Should be able to fire arbitrary statements. Access management still needs to be handled. One can’t ask Amazon and Barnes & Noble the same question, the same way.

People don’t care about new standards for coordination, etc. They want to get from A to B. Web Services are not an end, they’re a means. “The end is not to worry about plumbing,” it’s about getting the job done. Vendors should stop lying to the customers. “Modern-day charlatans” are pitching snake oil on how everything works like magic.

We need to start thinking about what a URL means. “We don’t know how to boil the ocean.” We can’t allow queries across distributed data.

Three methods to get data in a “mobilized data model”

Synchronization

Bosworth uses a RIM Blackberry 3 to 4 days a week. “Most of the time, I’m perfectly happy doing this.” But round-trips are expensive, and that makes browsing suck.

Queries

“I don’t know how we’re ever going to truly optimize these queries.”

Programming

BEA has XMLBeans, open-source API for reading and writing via XPath. It doesn’t navigate, but that’s being fixed. They need a data model to do that, to find out information that’s not clearly available.

Presentation is across a web, not across a document. It can pull data from multiple places to organize it, which is something XSLT doesn’t do well. Need to be able to work with services in context.

Service guys are not asking for WS-Transactions or Coordination, they want SOAP cookies (sounds like what my babysitter gave me when I swore -ed.). They want information models to work with, events to fire, and reliability. Bosworth mentions a company that spent $40B in IT over the last decade, and discovered that they didn’t get much, because things don’t work together.

Mobility needs to catch up. A number of wireless technologies are coming along, and need to work asynchronously. (The slide he showed looked very much like the syncing sessions I saw at PalmSource three and four years ago.) Synchronization will need to follow good role models like Blackberry. SyncML already exists, but should be upgraded to reflect this reality.

Bosworth got in a couple good jabs at Microsoft at the end: one on the closed release of InfoPath, and one on freezing the browser. He mentioned blogs (and RSS, indirectly) as a sign of new life in browsing data.