Re: 1.5: Discuss link characteristics?

In message <199702100113.RAA07568@boethius.eng.sun.com> bosak@atlantic-83.Eng.Sun.COM (Jon Bosak) writes:
>  5. link behavior: Links may have a wide variety of effects when
>     traversed, such as opening, closing, or scrolling windows or
>     panes; displaying the data from various termini in various ways;
>     testing, authenticating, or logging user and context information;
>     executing various programs. Ideally, link behavior should be

      ^^^^^  see below ^^^^^^^^^

>     determined by a semantic specification based on link types,
>     pointer roles, user circumstances, and other factors; just as
>     element formatting is determined by a stylesheet based on element
>     type, context, and other factors. It is recognized that there is
>     significant overlap between the areas of link formatting and link
>     behavior.
> The list of things that need to be specified is very helpful, but I
> don't think that the division between formatting and behavior is quite
> right yet.  For one thing, it seems clear to me that inline
> presentation without user action belongs to behavior rather than
> formatting.

I would be very keen to separate 'formatting' and 'behaviour' completely
if it were possible.  Most of the discussion here has (rightly) concentrated
on what happens when a document reaches a client where there is a human
involved.  Words like 'click', 'open a window', etc. are frequent.  May I
focus on 'executing various programs' and put ourselves in the position
of a robot who receives an XML-link document.  

I might think about database entries in XML which were automatically and 
randomly sent over the WWW. A robot might be expected to parse them
and their link structure, and take defined action (e.g. preparing a personal
database).  Presentation and formatting are not relevant to this process,
but the 'behaviour' of links might be critical.  For example, a link might
carry the semantics of visiting the resource at the other end and
abstracting that, possibly recursively.  The author might wish to vary the
'behaviour' as might the robot's owner.  An e-mail address/link might have a 
'behaviour' involving sending a message to the recipient automatically.

I get the implication that this behaviour can be completely determined
by stylesheets.  If so, the stylesheets need to carry 'behaviour' in a
way that can be implemented as a robot action, rather than a document
transformation, or a rendering.

I think of XML(SGML) documents as mapping onto Java objects.  Formally
(I think this is right) SGML documents carry only part of the content of an 
object as they lack state and methods.  I am not concerned about state
(I assume it could be added via XML if required?) but I believe that for
my purposes, precise behaviour can only be distributed via code (in my
case Java classes and methods).  Classes can be associated with Elements
(which is what I do at present) and in principle different variations
of those classes could be chosen for different 'styles' or 'behaviours.
I assume that something like that will also be possible for
LINK, but it will make it easier if the 'behaviour' is not inextricably
mixed with presentation.


Peter Murray-Rust, (domestic net connection)
Virtual School of Molecular Sciences, Nottingham University, UK