- From: Arjun Ray <aray@q2.net>
- Date: Mon, 31 Jan 2000 20:26:58 -0500 (EST)
- To: www-html@w3c.org
On Mon, 31 Jan 2000, Russell Steven Shawn O'Connor wrote: > I agree that this would be ideal, but I have an itch telling me > that the best solution somehow lies with the use of NOTATION. > But I'm not quite sure what that is. Get with the program, dude! NOTATIONs are SGML-geekery, and therefore to be avoided like the plague! But you're right. NOTATIONs play an important part. Unfortunately, there's still a problem - something that spans all of SGML. SGML has no general mechanism to bind two notations (or two entities) together. For instance, an ATTLIST for a NOTATION can't have attributes with ENTITY or NOTATION declared values (the obvious way to do it.) This restriction was to prevent potentially unbounded chains of entity dependencies - which ruled out the possibility of immediate dependency too, unfortunately. Removing this restriction didn't make it to the WebSGML TC either. This is important for <object> because it needs two NOTATIONs - one to describe the object (i.e. the program implementing the notation) and another to describe its inputs. The normal SGML kludge was to do the latter only, and conflate the NOTATION for the input with the program that groks it. That's fine if instantiating the program could be left to system configuration, but not at all OK when it comes to having to describe the program itself (or how to get it). So, there's a level of kludgery involved anyway (the problem is much worse in XML, which doesn't have data attributes) - at least one of the two ATTLISTs (for what are in principle NOTATIONs) will have to be for an ELEMENT instead. Since the properties of both code and data will be <object> specific, this means one set of "arbitrary" elements and/or attributes in the document instance. We *might* solve that problem via architectures (i.e. require that "definitional" element only to match a form) - but the W3C won't have that either. So, it's down to horsetrading and infighting over which and how many attributes - 42?:) - for <object>. Arjun
Received on Monday, 31 January 2000 20:14:18 UTC