- From: Clark C. Evans <cce@clarkevans.com>
- Date: Thu, 20 Jul 2000 18:00:22 -0400 (EDT)
- To: Jonathan Marsh <jmarsh@microsoft.com>
- cc: "'www-xml-xinclude-comments@w3.org'" <www-xml-xinclude-comments@w3.org>
Let me try and patch up your questions/concerns. On Thu, 20 Jul 2000, Jonathan Marsh wrote: > 3) The default view is that the included tree is not > transparent to inclusion. How about "xinclude:strip-after-include" where the default value is 'default'? (just like the xml:space attribute?) > I prefer something like an infoset addition that provides > pre-inclusion information for those processors sophisticated > enough to expose it. Yes, but this would cause yet-another dependency between specifications. If a new axis was added to the infoset, then XSLT would have be modified, etc. Yuck. > The simplest processors would still be completely transparent. Yes, this is the goal. However, the primary advantage of XInclude over Entities, as I see it, is that it uses the XML syntax -- so that processors like XSLT can be aware of them. As far as transparency is concerned, this is what namespaces are tasked with solving! no? > 1) What if I include an empty element? Hmm. How about "xinclude:state" which can take on the values: active - processors should include when this is set (the default) included - the inclusion has already occurred (only when strip='false') inactive - skip inclusion for now... (good for XSLT) > 2) Merging attribute lists looks like a significant complication. Yes, kinda ugly. However, I really *do* like the idea. It provides for inheritance of sorts. But, it might not be worth the complexity. > I agree that it seems wrong to throw out the user's element and all its > attributes and even content just because of the presence of a couple of > attributes. I am leaning back toward the element syntax of a previous draft > for this reason among others. Yes, it just kinda feels wrong. Hope this helps... Clark
Received on Thursday, 20 July 2000 17:48:25 UTC