- From: Joe English <jenglish@crl.com>
- Date: Mon, 21 Apr 1997 12:11:20 -0700
- To: w3c-sgml-wg@w3.org
Tim Bray <tbray@textuality.com> wrote: > To summarize: I proposed that XML processors be required to stop > passing data (other than error notifications) to applications after the > first violation of well-formedness. This proposal does sound appealing, but I agree with the others that it's too draconian. Here's what I would like to see as a requirement: 1. An XML parser _must_ report the first error it detects to the application, as soon as the error is detected. 2. XML user agents _should_ notify the user of any errors encountered in an XML document. [ By "parser" I mean a chunk of code which translates XML into another form suitable for manipulation by some other chunk of code; the "application" is the other chunk of code which does the actual processing. By "user agent" I mean an application that renders XML in a form suitable for human consumption. ] Requirement (1) is similar to the "C" flag that SGMLS (fails to) generate at the end of a (non)conforming document, except that it would be reported earlier. This would leave it up to the application to decide what to do with bad XML. Mission-critical applications (like financial transactions) and applications that are highly sensitive to correct input (like XML-LINK) could decide to stop processing immediately, while things like display engines could plow ahead. As for requirement (2): most HTML is deemed correct by its creators if it passes the "looks good in Netscape" test. A blinking red light in the status bar saying "Bad XML" would go a long way towards preventing ill-formed documents from ever being published; no matter how good the error recovery in the display engine is, that blinking red light would make bad documents fail the "looks good" test. Of course, as with Terry's proposal, parsers would be free to report as many errors in as much detail as they like; that's a quality-of-implementation issue. The main point is that parsers must inform applications, and applications should inform users, if the input is no good. --Joe English jenglish@crl.com
Received on Monday, 21 April 1997 15:13:31 UTC