Saturday, April 28, 2007

Declarative Systems and the Cambrian Explosion

I have been reading Carl Zimmer's book Evolution, The Triumph of an Idea and I have been startled about the similarities between the the Cambrian Explosion and the growth of declarative languages.

Just to recap (if you have not read the Wikipedia article yet) here is a quick summary. For the first 3.5 billions years, life on earth was mostly single celled animals. Then about 550 million years ago, the number of new multicellular complex-body-plan animals exploded. There are two salient reasons. First, we needed oxygen to breath. Before there was oxygen in the atmosphere respiration was not easy. Oxygen was a foundation for complex body plans on land. Second, life needed a genetic "tool kit" to build complex animal shapes. We needed the HOX genes for building complex body plans. After that happened, complex life forms did explode.

I am seeing the same thing appear in declarative languages. Declarative languages need foundations also. We needed ways for people to quickly create new languages (e.g. GUI XML Schema editors like XMLSpy) and ways to build consensus around the semantics of these new languages (wiki's, social networks, folkonomies and voting).

And next we need the declarative systems tool kits. Right now we have started to develop tools to allow XML Schemas to be quickly transformed into GUI XForms editors and we have XML databases that eliminated the need for complex middle tiers. We can put many of these tools in the hands of non-programmers and allow them to create domain-specific vocabularies and rule-based systems with GUI editors.

And when those tools mature we will see an explosion of declarative languages similar to the explosion of life forms 550 million years ago.

So what is your prediction? 1 year, 3 years, 5 years or 10 years from now? I think a small group of people could build this toolkit without millions of dollars of venture capital. There may even be open-source tools that do this in a few years. But I think the key is to combine the best of metadata registries, XML Schemas, transforms, XForms, XQuery and XML databases into a holistic declarative system.

And from here the many NEW ideas will evolve.

Saturday, April 14, 2007

XForms and the Chasm: What XForms Need to Become a Mainstream Technology

Geoff Moore wrote a very insightful book on technology adoption called “Crossing the Chasm”. If you study the psychology of technology adoption, you find that people and organization fall in a bell curve. About 2% are innovators, 14% are early adopters, 33% early majority, 33% late majority and 15% are laggards. This is all standard stuff out of basic marketing-psychology textbooks. Each group has different reasons for using (or not using) a specific product.

Moore's insight was that many new products never move from early adopters to early majority. He called this region “The Chasm”. I have found that the same rules applyto new standards and new languages.

Moore did a lot of analysis on why some new technologies don’t make it across the chasm despite their merits or why some standards are adopted over others. Hisbiggest finding is that they were just too hard to use by non-techies. They required too much work to setup, install or learn. Ease-of-use was a principal factor in many technologies not being quickly adopted by organizations that did not want to use technology to differentiate their organization. Early Majority customers also tend to purchase new technology not based on how they work but who is using them in their industry. They purchase on references.

New technologies usually need a focused approach in a specific industry to be adopted by an early majority. If an organization can target a specific vertical industry and provide a more complete solution for that industry, once a few reference accounts are set up the rest of the industry can see there are proven case studies.

Another factor is the need of the early majority for complete solutions. Innovators and Early adopters are willing to glue together new technologies into their existing fabric of solutions. They dedicate time, staff and money to do this. If their XForms sends XML data over REST interfaces and their databases use JDBC they hire programmers to write middle-tier applications.

But early majority buyers can’t justify that additional cost. They just want a solution to work out-of-the-box. They are OK paying for a week for training for their BAs, but they don’t want to write and maintain complex applications.

Another thing that I have been searching for are metaphors that are understandable by non-programmers. Since the high-level decision makers are usually non-technical their ability to make a decision is usually based on their trust of the developers or their understanding based on metaphors.

The Language Translation Metaphor I find that if you can send your data from an XForms application directly to a native XML database in XML format that there is no need for middle-tier objects or shredding into relational databases. No translation is involved. One of the ways to demonstrate how machine-based language translation is not always semantically precise is to try an automatic translation program. Convert a statement from a native language into a foreign language and then translate it back. For example try this using Google Language Tools at http://www.google.com/language_tools

Original Phrase in English: The quick brown fox jumped over the lazy dog.

Google Translation to Spanish El zorro marrón rápido saltó sobre el perro perezoso.

Spanish Translation Back to English The fast brown fox jumped on the sluggish dog.

Although Google translate does a pretty good job, quick was changed to fast, over was changed to on, and lazy was changed to sluggish. If you study semantics, you can find lots of problems trying to communicate with close approximations. Note the original implicit business rule of don’t repeat a letter was also lost. We have many implicit business rules in complex systems and translation of these to other formats is always problematic for software developers.

Saving XML using an object-middle tier is one translation. Saving objects to a relational database is a second translation. Fetching data from a database using SQL is a third translation. Converting tabular data into objects is a fourth translation and converting objects back to XML is a fifth transformation. There are five translations, not just two.

Not that XForms solutions today are not without translation. In our process we are planning to convert XML Schemas into XForms as well as initial instance data. We need to also import leaf-level definitions from a metadata registry. But the nice thing about this process is that we can validate the submitted data directly against the original XML Schema that was used to generate the XForms. An elegant self-checking pattern.

All of these transforms are possible. But all have to be maintained each time you change your XML Schema. Right now we also need great tools that translate XML Schemas directly into XForms. And these need to retain the fidelity of the original intent of the subject matter experts.

So perhaps deploying applications with native XML databases is a critical part of the XForms value proposition.

Moore also noted that both early late majority buyers would be more willing to use a technology if it was buried deeply in a complete solution. That way the IT managers don’t have to sell the technology to management, just the solution. Similarly, once third party developers start building accounting or complete solutions for vertical industry using XForms the buyer will not need to know or care how the solution is being delivered.

Friday, April 06, 2007

Metaphors for Declarative Systems

I am working on my presentation for the 2007 semantic technology conference. My topic is the semantics of declarative languages.

I am attempting to present to business strategists that may not appreciate how powerful the new XForms and XQuery standards are and how entire rich user experience web 2.0 applications can be built using these tools as long as the tools have RESTful interfaces.

My first consideration is to find the right metaphors. I am working with the evolution metaphor and the puzzle metaphor right now and it seems to be working well.

I have also done some research on Domain Specific Languages and I see that they have also used the evolution metaphor but I find that they are focused on building small languages that are specific to a group or project and they don't really look at the external semantics of the problem.

Please write me if you are also trying to explain the business benefits of either declarative languages to a non-technical audience and if you have found metaphors that get the key points across.


Thanks! - Dan