ISSUE-48: POSTPONED: information on best practices for tools builders
tool best practices
POSTPONED: information on best practices for tools builders
- State:
- CLOSED
- Product:
- Raised by:
- Peter Patel-Schneider
- Opened on:
- 2007-10-28
- Description:
- There already has been discussion on serialization of ontologies, http://lists.w3.org/Archives/Public/public-owl-wg/2007Oct/0307.html, as well there are other issues that are not really normative parts of the specification, like keeping track of names in QName (or CURIE) form instead of turning them into full IRIs.
It probably would be a good idea to put all this in a separate, informative section (of SS&FS?), partly to keep it separate from normative stuff. - Related Actions Items:
- No related actions
- Related emails:
- ISSUE-48 (tool best practices): REPORTED: information on best practices for tools builders (from sysbot+tracker@w3.org on 2007-10-28)
Related notes:
While I think some best practices for tool builder stuff would be quite valuable, that is distinct, I think, from whether we spec tigher serializations. There's no reason why we can't, even for loose serializations, define a canonical form. Whether a tool produces only canonical form (yes!) or only consumes it (no, at least for RDF/XML) is, indeed, a best practice issue.
Another way of thinking of it is that it induces a new class of tool: canonicalizers, where as QName preservation is a constraint *on* the tool. (QName preservation seems to be a reasonable contraint on a canonicalizer, if it starts from a source with QName information.) I guess you can just go with a relational form of canonicalization, which makes the canonical form the canonical form of another document (for example). These distinctions aren't hard and fast.
Out of scope (but see ISSUE-14).
Ian Horrocks, 1 Dec 2007, 19:49:29Display change log
