- From: Henri Sivonen <hsivonen@iki.fi>
- Date: Wed, 31 Aug 2005 18:51:06 +0300
- To: Bjoern Hoehrmann <derhoermi@gmx.net>
- Cc: www-validator@w3.org
On Aug 31, 2005, at 03:13, Bjoern Hoehrmann wrote: > * Henri Sivonen wrote: >> * Doctype and schemaLocation are design bugs, because the document >> that is being assessed sets its own rules instead of the document >> being >> assessed being subject to rules set by the person assessing the >> technical quality of the document. > > That's both good and bad. "Versioning" is a reality and its often > possible to determine the defnite rules with respect to standards > compliance solely based on the document. I suppose there is merit to asking "What does this look like?" and then asking immediately if it is valid according to the answer to the previous question if the validator tells the user what schema it is validating against is and the schema is still provided by the user or the validation service and the document cannot covertly inject rules of its own. > There may be stricter > rules imposed at a higher level, but it is not feasible to require > users to select a specific schema each and every time they want > to validate something. It can be a user interface annoyance, yes. That's why I made to settings bookmarkable in my validation service. The main issue is whether it is useful to know that a document is valid according to some schema (for each XML document there exists [in the mathematical sense of existence] a schema according to which the document would be valid) or whether it is useful to know that the document validates according to a particular schema. >> * RELAX NG (Schematron too) fixes the bug: the RELAX NG validation >> process takes two independent inputs: the document and the schema. > > This is no different from DTDs really, that most tools are only > able to tell whether an SGML or XML document is "valid" per the > rules in the DTD and the internal subset and such is an API bug; > not all toolkits have this bug. It is substantially different. With DTDs (as specified--I know there are tools that allow different things) the validation process takes one input (with possible includes). >> * If validator.w3.org adds RELAX NG support in response to what the >> SVG WG is doing, it would be nice if the two-input nature of the >> process was preserved and not regressed to the DTD ways by using >> heuristic schema association. > > You would likely be able to use the Validator as a general purpose > RELAX NG Validator but it's unlikely you will be required to select > the schema for each validation. It's not just the authors who want > to check a document, it's also customers of web design companies > who want to check whether their produce proper code, and not all > customers are aware of all the technical details relevant here. So valid FooML is good enough if valid SVG 1.2 was ordered? -- Henri Sivonen hsivonen@iki.fi http://hsivonen.iki.fi/
Received on Wednesday, 31 August 2005 15:51:20 UTC