>At 01:19 PM 10/9/98 -0400, Bill la Forge wrote:
>>Two concerns here:
>>
>> 1. There should be a way to cascade whole Bind documents,
>> not just individual entries, as well as adopting a
>> first-encountered
>> entry rule. This would allow one Bind to override another,
>> dropping inappropriate items under an entry.
>
>So, just as a stylesheet could be designed for a particular DTD,
>so too could a Bindings document. Or a Bindings document might
>apply to a "library" of element/attribute definitions.
A generic Bind document may be bound to a particular
DTD. But an application specific one with Java class bindings
would probably not be--as a DTD may be usable by more that
one program, each with its own element/class bindings.
>I think that you are also underscoring a need for delegation.
I don't follow.
>I'm not yet sure whether I agree with the first-encountered rule.
>We need to examine the cost of not allowing bindings to be cumulative.
Bindings can be cumulative by making use of the cascading link
internal to the entry. (This is definately getting ahead of the discussion--
I shouldn't have raised such a detail.)
>
>
>>
>> 2. Binding an entry to a class should be accomplished without
>> recourse to a link. This would allow for light-weight
bindings.
>
>That would be one way to look at it. Another way to keep it
>fairly light-weight is to allow that a set of Bindings might
>be contained in the same document. Thus, we would maintain
>a link mechanism is all cases, but allow for the equivalent
>of a fragment identifier (#) to say that te Binding is located
>in the self-same document.
That does get you more than half way there.
But if the binding for an entry is nothing more than a few
data items, it seems more appropriate simply to
include them right there in the entry.
I think the key here is that we are dealing here with
information specific to a application. So bindings would
likely be defined mostly at the top level and rarely at the
level where the Bind document is dealing with generic information
applicable to all uses of a particular markup language.
Bill