XML 2005: Document Formats

As chair of this session, it seems appropriate I provide personal notes. First, let me compliment all attendees on the tremendous and evident respect they brought to the discussion and each other. You are clearly leaders and deservedly so. While none, I think, believed decisions could be made in the town hall, your seeking common understanding in respectful debate was a marvel to experience. Thank you.

The following are questions that were in my opinion, pivotal in the discussion with my personal opinions about the issues arising from them.

1. What is open?

A clear consensus for this question did not emerge. All sides have valid arguments for the technologies they sell, specify and standardize. Questions of the open and closed market models resolve to company choice.

Opinion: The participation agreements of the consortia make adoption and participation separable. While there is an emerging consensus on the qualities of openness, their is no policy or means of enforcement without a clear means of identification. It may be sufficient to say that open is best defined in accordance with the policies of the publishing authority such as OASIS or the W3C given an ontology of concepts for that term and the provability of assertions of duty in the records of authority for the obligated parties. It will be well for the market if such ontologies are sharable among consortia and government alike. This is a significant challenge.

NOTE: Tim Bray made the valid point that the participants in the discussion are not legal experts. IANAL.

Opinion: It is clear that all sides have market ambitions. Mutual interests are likely to converge around costs of sustaining loss leader technologies. A specification represents opportunity; a standard represents market reality as costs to producer and consumer. The advantage of a standard to the vendor is that development costs can be shared and components are interoperable. The advantage to the consumer is that costs can be controlled. A clear shared interest is cost control.

2. What is open enough?

Given item one, there is no consensus here. However, this is where the most progress is being made. XML provides the opportunity to create a common markup for word processing document types and much technology is being provided for systems to share formats. At issue is sharing semantics such that the rendering and behavioral qualities are predictable and reliable for the end user as far as possible.

Opinion: The concern expressed by Microsoft was early adoption of standards such as XSD are costly and if not well liked, contribute to premature obsolescence. In my opinion, this is a risk all early adopters take. The vendor is expected to evaluate a specification and the process of its development. Thus, this is not simple a question of open enough, but also, good enough. In the market, I expect consumers to understand the relative development maturity of software and to choose wisely. It works for cars and other consumer goods. All sides assume this risk.

The concern expressed by open software advocates has both practice and egalitarian appeals. They assert that open specifications developed by all of the market vendors provide the best software, a position sometimes known as, the wisdom of crowds. However, XSD. It is a workable technology, but is it the best or good enough? So in fairness, this wisdom does not always work as advertised. If the best advice is to create as few languages as possible, parties must create very good ones or this strategy saddles the market with mediocrity.

Again, this is a risk of early adopters. RelaxNG is a viable alternative. The problem of the market is sales cycle. A vendor does not see a customer again for some time. The question of standardization can vary by application. The strategy of specification over standardization is to give the market time to decide. Given this, for specification of new products, it is prudent to standardize as little as possible to enable decisions to be made more granularly and timely.

The winning strategy is found in reducing the complexity of the application itself, such that the end user is receiving only those components clearly needed by role and task or preference. This is where the question for egalitarian positions is: given the disadvantage accrues to the vendor with the most legacy, how does the vendor adopt an open specification if the immediate results break fiduciary duty to the company ownership?

But is this question that cut and dry?

Metaphorically, legacy is to market what mass is to momentum and distance traveled over time. If one suddenly cuts the mass, the distance over time can beat the speed of light. An example as Matt Fuchs pointed out over breakfast is XML itself. The task of refactoring SGML for web applications cut mass. Then the market achieved momentum and distance.

So, if the answer to what is open enough is, the market chooses and what the market chooses becomes the standard, then a specification must achieve enough mass in terms of users to achieve momentum. The conundrum is the vendor of a product with sufficient legacy releases a large number of customers as well as software support when legacy is terminated. So the vendor will wait until the software components and the language achieve a point of inflection.

Analysis of Outcome: The open document community must keep up the pressure. Microsoft will implement openDoc as part of a strategy to move their customers from their loss leaders to lower cost software that enables a switch to service systems. The question is timing and the complaint is that this is not an open market. This is not as open a market as some need or think more provident to more companies, but it is an opportunity to do this better than the current marjority market leader.

Because a standard becomes a game model where the evolutionary stable strategy is a Nash equilibrium, it is best to defer standardization until need is clear to all parties. Disruption isn't always good, but winning need not be defined as owning.