- From: Michael Sperberg-McQueen <U35395@UICVM.UIC.EDU>
- Date: Mon, 28 Apr 97 14:14:56 CDT
- To: W3C SGML Working Group <w3c-sgml-wg@w3.org>
On Mon, 28 Apr 1997 05:14:15 -0400 Tim Bray said: >I agree that if XML were to be used in the current single-purpse way that >HTML is, i.e. exclusively for browsing, then error recovery would be >desirable. But I do not, and so remain convinced that we have to >make a strong statement in order to achieve a violent break with this >one particularly pernicious cultural characteristic of the Web. Why does the strong statement have to take the form of *forbidding* error recovery? Why is prescribing an inappropriate error recovery all right in the application area (browsing) that is already well established and not all right in the application areas we have not yet widely implemented but think we want to? What is wrong with allowing applications and XML processors to decide for themselves whether it makes sense to attempt to recover from errors? >I think that Rick's use of the term "sudden death" to describe what >I'm asking for is perfectly appropriate. > >A request: could the product vendors and large-scale information >providers on this group please put their hands up and say what they would >like to see in this area? Reason I ask is, a lot of the arguments against >sudden death have amounted to "oh, the vendors would never support it." I'll take this on faith; I haven't seen any such arguments. The only vendor-related arguments I've noticed are those implicit in the Draconian position. Tim seems to be assuming silently that vendors are too stupid to notice that some applications require failure on error detection, and too weak-kneed to implement it if it's not forced upon them and they're not guaranteed that their competitors can't attempt error recovery either. So I want to ask the implementors another question: is it enough that an XML processor be required to issue error messages, and allowed to treat errors as fatal errors? Or is it essential that XML processors be *required* by the spec to treat WF errors as fatal errors? If the latter, can you say why? -C. M. Sperberg-McQueen
Received on Monday, 28 April 1997 15:27:12 UTC