- From: len bullard <cbullard@hiwaay.net>
- Date: Fri, 09 May 1997 19:41:01 -0500
- To: Gavin Nicol <gtn@eps.inso.com>
- CC: Todd.Freter@Eng.Sun.COM, bsmith@atlantic-82.Eng.Sun.COM, w3c-sgml-wg@w3.org
Gavin Nicol wrote: > > >Moreover, neither the current state of the Internet nor the availability of > >other protocols or data representations invalidate Bill's argument, which is > >about XML, fault-tolerant applications, and how the current ERB decision makes > >XML beg off from processing them. > > I guess if the only tool you have is a hammer, everything looks like > a nail.... > > For mission critical applications, realtime constraints and data > redundancy are generally required. This to me implies higher level > protocols than those codified in the data. XML doesn't solve this > problem, and shouldn't pretend to. Systems of the type Bill and Jon describe do not use XML for real-time data collections. They use it to move the collection into other database modules. For every sensor application used for safety control systems there is a records management system that collects and broadcasts that information in filtered forms to different collection agencies. These agencies use that collection to extract information for feedback to decision systems. The decision systems tolerance for error depends on the relationship of time from incident to decision and decision to dispatch. I think the problem here is drawing the line between what conformance means to implementation, or simply, what is standard and what is application. Who sorts errors? How far do we as application implementors go with the spec and when does it become our decision to determine application design? Some applications only want raw data in errorless form. Some applications can handle errors. The mission-critical approach emerges from time characteristics that determine where classes of errors are trapped by tools and tool users. Lee Quin stated the boundaries very well. lwn
Received on Friday, 9 May 1997 20:41:19 UTC