Re: Best Practices Document on Caching XML Schemata

Ted Guild wrote:

> The purpose would be to increase awareness of the
> situation and subsequently pressure on resolution with
> library and software developers.

It is clear that a significant number of the requests W3C's servers have received have been generated by software built with the .NET System.Xml or MSXML libraries. Several of us at Microsoft have, accordingly, been communicating with Ted Guild to discuss potential solutions in light of the challenges W3C faces in hosting the standard schemas. Unfortunately, currently known fixes for this problem would probably be more substantial than we would normally put into a service pack, and features for the next System.Xml release in .NET 3.5 are already locked down.

We are interested in the TAG's perspective on this issue, and welcome the opportunity for a public discussion of the challenging issues that surround efficient DTD/schema resolving. This situation touches on some famously hard problems, and we could use all the brainpower that can be applied to finding a globally viable solution.

A few preliminary (and informal) questions and thoughts:

1) What guidance do the Web Architecture principles offer here? One great strength of the Web is that DTDs and schemas are accessed with the same operators and semantics as any other resource, but this leads to a bit of a dilemma: Where in the XML / HTTP / URI stack would reside the knowledge that DTDs and schemas are resources with fairly static representations and thus can be accessed from a persistent cache? Likewise, where in the overall Web infrastructure should these DTDs/schemas be cached?

2) To what extent does the TAG believe that a single Best Practice for schema/DTD caching would apply across client, middleware, and server environments? If not, how might library developers implement different practices for the same library used in different scenarios?

3) People with a deep understanding of XML and the Web technologies can be expected to effectively implement the various solutions outlined in Norm Walsh's article (linked to in the original message), balancing the pros and cons of each. Library and software developers, however, do not have the luxury of assuming that their customers have this level of sophistication. Accordingly, a highly configurable solution that library developers could implement, and that could meet W3C's needs, might not be viable if it significantly increased customers' deployment costs or library suppliers' support costs.

4) We are mindful that this might be primarily an economic dilemma whereby the W3C, having successfully worked to increase the number of applications that access the standard DTDs and schemas, does not realize any additional revenue to offset the costs of increasing traffic.

Michael Champion

Received on Wednesday, 1 August 2007 21:28:41 UTC