Re: elm alternate linking proposal

At 03:31 PM 2/5/97 -0500, Eve L. Maler wrote:
>I've been working on an alternate proposal...

There is nothing in Eve's proposal (hereinafter ELP) with which I can bring 
myself to disagree violently.

I *think* that the design choices in ELP can all be decided on within the
question/voting framework I published last week; but it is certainly a good
idea for Eve to have trotted out the whole ELP framework so that we can
get the context for the positions she will doubtless be endorsing.

A couple of detailed design comments:

 - In general I support less-is-more.  But I doubt that we're going to be
   able to get away with doing *nothing* on the issues of explainers, 
   terminus-roles, and behavior/formatting metadata.  It is absolutely
   dead certain that people are going to put this stuff in their links;
   at the very least we should probably give them a standard place to
   put it.  On the other hand, I'm starting to be convinced that locsrc
   is more trouble than it's worth.
 - I'm not sure we want to embrace the whole power of archforms to the
   extent of allowing attribute name overrides; not that another level
   of indirection isn't always a good idea, it just makes it a bit
   harder to explain.  I think in XML, we have gotten good mileage out
   of nuking everything that was in the slightest hard to explain.  It's
   *so* easy to explain, "once you've figured out that this thing is an
   MLINK, then here are the attributes and here's what they mean."  Also
   the tokenized attribute saying "consider this attribute to be that
   attribute" feels kind of klunky to me.

 - A design principle I realized I had used implicitly without writing it
   down was: you should be able to extract all the linking information with
   just an XML processor and nothing else.  This is why in the initial
   draft proposal, there is no case in which you have to tokenize 
   an attribute string.  This is just an invitation to problems in the
   i18n arena; and furthermore, I've always thought that the base principle
   of SGML was that you should take things that are supposed to be 
   separate, and separate them with markup.  This explains my discomfort
   with the ELP "xmlnames" attribute mechanism.

Cheers, Tim Bray
tbray@textuality.com http://www.textuality.com/ +1-604-708-9592