Re: XHTML & hyperlinking opinions (long, sorry)

Tim Bray writes:

>> I think reasonable disagreements are of the 
>> form "no, that's not a valuable thing to add" 

I agree that's the key issue, but suggest a reformulation:

        "I think reasonable disagreements 
        are of the form 'no, trading off the 
        value gained against whatever 
        drawbacks, it's not on balance a 
        good thing to add.'" 

Xlink presumably has value in at least some situations.  I think the 
question is:  how often, how much, and how does that value compare to 
whatever downsides it might have?  We need those answers in general and 
for XLink-for-XHTML in particular. 

We also have to be clear on what we're advertising as the value gained 
from any particular proposal.  What I think I've heard argued for as the 
benefit of XLink includes at least the following 3 points (I've thrown in 
personal opinions on a few, but they're not central to my main suggestion 
which is: list the proposed benefits& drawbacks, and do a dispassionate 
pro/con analysis):

----Possible benefits of XLink for XML ---------
------in general and XHTML in particular--------

1. A linking mechanism that is common across all (or most) vocabularies 
and applicable to all uses in those vocabularies, so that it can be 
recognized independent of context. 

This is just as all attributes are represented uniformly, and we can 
therefore build tools that work on any attribute in any vocabulary. The 
claim would be that  we need the same uniformity for links, presumably so 
we can build things like generalized, vocabulary-independent link 
manipulation tools.

(I happen not to be convinced that this is in general practical, because I 
think links have a semantic that's not in general trustworthy without 
knowing the application context.  For example, see my earlier posting on 
undo lists containing deleted data in compound document structures [1]. 
You don't really understand a link, I claim, if it's in a structure that 
the application uses to represent deleted data.  Or conversely, a 
construction like XLink might be restricted to use in situations where 
context doesn't matter, implying that when I move deleted text to the undo 
list, I have to change from XLink to something else (so that generalized 
tools won't think this document is still making a link.)  So, I remain 
suspicious of the need for or practicality of a common way to encode all 
links.)

2. A richer linking mechanism with multi-way links, third party links etc. 
 This seems to have value, at least in principle, and figuring out how to 
do it once and reusing that insight where appropriate seems to make sense. 
 Whether the value of trying to actually get to the level of common 
serializations of these constructs in application-specific vocabularies 
outweighs the complexity and inconvenience, I'm not sure. 

3.  A generalized means of providing presentation hints, what I believe 
XLink calls behavior attributes.  (I have some nervousness about these 
too, in that I see much (not all) XML as being on the model side of a 
model/view architecture.  Insofar as these attributes encourage embedding 
of presentation hints in the model, I get a bit nervous.  That said, XLink 
layers them quite well, and makes them optional, and they certainly are 
potentially applicable to presentation-oriented vocabularies such as 
XHTML.  Still, I'm not sure how much value there is in generalizing them 
across vocabularies: probably some, but I don't see it as broadly 
applicable to all XML, which is what I thought was advertised for XLink.)

-------------------------------------------------

So:  for Xlink, I think we have to answer Tim's question (as reformulated) 
with respect to the three purported benefits above.  I personally don't 
have a strong opinion pro or con on XLink, but am interested in seeing the 
questions stated clearly so we can get to an answer.  Maybe this is a 
piece of the formulation?

Noah

[1] http://lists.w3.org/Archives/Public/www-tag/2002Sep/0178.html
------------------------------------------------------------------
Noah Mendelsohn                              Voice: 1-617-693-4036
IBM Corporation                                Fax: 1-617-693-8676
One Rogers Street
Cambridge, MA 02142
------------------------------------------------------------------

Received on Friday, 4 October 2002 23:13:45 UTC