- From: Henri Sivonen <hsivonen@iki.fi>
- Date: Sun, 11 May 2003 20:25:36 +0300
- To: www-html@w3.org
On Sunday, May 11, 2003, at 06:27 Europe/Helsinki, Jelks Cabaniss wrote: > I don't think an XHTML-aware UA should be downloading the XHTML > entities > when it's dealing with the fixed vocabulary it already knows about! An application (in the software program sense) using an XML processor to parse a document marked up in an XML-based language needs to know about the elements and attributes but the entity names should be of no concern to the application. The character entities should be dealt with at the XML processor level and a conformant XML processor only knows about lt, gt, quot, apos and amp a priori. If external entities are to be processed, the alternative to downloading is maintaining a local DTD catalog for a finite set of public ids. This approach breaks when someone comes up with a homegrown combination of the XHTML modules and uses a public id that isn't well-known. Mozilla a DTD catalog, but it cheats: The DTDs in Mozilla's catalog are so heavily abridged that they only contain the character entity declarations. The character entities work as if Mozilla was reading the real DTD but other stuff like attribute defaulting and the #IDness of attributes don't work. The DTDs are *huge* compared to the content length of usual Web pages. Parsing an entire DTD every time a document is loaded makes no sense in an interactive application. The concepts of XML well-formedness and non-validating parsers were supposed to relieve interactive applications from dealing with external DTD subsets. -- Henri Sivonen hsivonen@iki.fi http://www.iki.fi/hsivonen/
Received on Sunday, 11 May 2003 13:25:45 UTC