- From: Henry S. Thompson <ht@cogsci.ed.ac.uk>
- Date: Tue, 11 Feb 97 16:05:04 GMT
- To: w3c-sgml-wg@w3.org
OK, it's REALLY time for back to basics. I'm getting close to a redraft of section 1, but here's the background for the (aspiring) rocket scientists, since I'm doing my best to write the redraft for the web heads. We've really got (wait for it) SIX sorts of things, types and tokens versus notation, description and thing itself, viz: notation description ding an sich type DTD fragment linkd class link class instance linkde linkd instance link instance You can start at the lower left and get the sense of this. We have the following link description element in an XML document, say <tm href="http://www.microsoft.com/">Windows 95</tm> This is a instance of a type specified by the following DTD fragment (apologies to Tim and Steve, not quite as per the existing document): <!element tm - - (#PCDATA)> <!attlist tm XHL CDATA #FIXED "TLINK" HREF CDATA #REQUIRED HRTYPE CDATA "URL" TROLE CDATA #FIXED "owner" SROLE CDATA #FIXED "owned"> The text is notation for a link description instance, whose internal representation in some object language we could notate as follows: [tlinkd name=tm source=[epd loc=<the A element itself> name=owned] target=[epd loc=[url ttype=http host=xyzzy.com doc='']] name=owner]] That's an instance of a (subclass of) link description, with values for various properties. The values of the source and target properties are themselves instances of endpoint descriptions. This in turn, in the context of a particular application, document and location thereof, might resolve to a link instance we could notate as follows: [link name=tm eps={[ep name=owned pt=*1], [ep name=owner pt=*2]}] where *1 and *2 are hard pointers to groves or grove fragments or whatever. Note that MANY different link descriptions might have given rise to this link, whereas the relationship between link description elements and link descriptions is much tighter. Link description types and link types are just the appropriate abstractions, with there being some structure to the family of link description types (i.e. at least we have tlinkd and ilinkd with some shared and some distinctive properties), but as far as I can see there is only one link type, since I don't grok directionality yet. Where is this headed? Well, here's a translation of old terms into new terms, easier first Old New terminus endpoint pointer endpoint description link-type link name pointer role endpoint name locator locator explainer gloss link link or link description or link description element All the terms 'link name', 'endpoint name', 'locator' and 'gloss' may be used informally to apply to the constituent parts of link descriptions and link description elements indiscriminately. Be patient, this all IS worth it, I promise. ht PS. Lest there be any confusion, I don't see this as an attempt to start over. I claim I'm engaged in a rational reconstruction of the draft on the table, with virtually no change to the fundamental content (Well, one, in that I don't see any way therein to give the pointer role (old name) for the implicit terminus (old name). In the above example, the attribute 'SROLE' (source role) gives the endpoint role for the (implicit) endpoint which is the link description element itself, while the 'TROLE' attribute (target role) gives the endpoint role of the endpoint explicitly described by the HRTYPE, HREF, etc. attributes.).
Received on Tuesday, 11 February 1997 11:05:21 UTC