Re: [uiml] Summary of issues raised by Takashi Endo and suggested UIML 4.0 changes

["Robbie Schaefer" <robbie@c-lab.de>]

Hi,

I got a chance to dive into the comments of Mr. Endo. And I see something 
which affects the complete DTD:

Principally, the variable "id"s (and in fact all "id"s we have) don't 
conflict in the DOM tree since they are not declared as real id's in the 
DTD. Each id attribute we defined is of the type 'NMTOKEN' not of type 'ID' 
so it is perfectly OK if it occurs multiple times in the DOM-tree. So with 
the current DTD of course getElementByID does not work, since there are no 
ids and implementors must therefore use other mechanisms.

As Jim stated in a previous mail, uniqueness of IDs is not useful for all 
elements, so we must take a careful look for which elements to set the type 
of the id-attribute to 'ID' instead of 'NMTOKEN'.

All the best,
Robbie.
--
   _/ Dr. Robbie Schaefer        _/ Phone: +49 5251 60-6107  _/
  _/ Visual Interactive Systems _/ Fax: +49 5251 60-6065    _/
 _/ C-LAB Fuerstenallee 11     _/                          _/
_/ D-33102 Paderborn          _/ URL: http://www.c-lab.de _/

----- Original Message ----- 
From: "Jim Helms" <jhelms@gmail.com>
To: "Robbie Schaefer" <robbie@c-lab.de>
Cc: <uiml@lists.oasis-open.org>
Sent: Monday, December 03, 2007 3:38 PM
Subject: Re: [uiml] Summary of issues raised by Takashi Endo and suggested 
UIML 4.0 changes


> Robbie, thanks for you input!  Additional comments below.
>
> On 12/3/07, Robbie Schaefer <robbie@c-lab.de> wrote:
>> > Suggested change: add an attribute to <variable> called "id-ref" that
>> > is used to reference a variable after it is declared.
>>
>> We had thought of this when designing DISL and the current sulution in 
>> the
>> UIML-DTD is that we have an additional attribute for the variable which
>> specifies wether it is used as a declaration or as a reference: reference
>> (true|false) "true".
>>
>> However if all agree that id-ref is the more proper/consitent solution
>> (since it is used in other UIML-elements in the same fashion) I could go
>> through the relevant parts in the spec and change the examples and
>> explanations.
>>
>> The point in favor of the "reference"-Attribute is that it is implicitly 
>> set
>> true, so that only once, for the declaration, a variable needs to state 
>> the
>> reference attribute, which IMHO adds to the readability of the
>> UIML-document.
>>
>> Anyway, I am open to both options.
>
> That is a good point.  I like the way DISL handles this, but we have
> two considerations: 1) UIML sets a precedent with the <property>
> element that establishes "name" as the way to reference an existing id
> without conflicting; and 2) the additional reference attribute still
> leaves us with the issue of having conflicting id's.  The DISL
> approach is very good for readability, but I think the problem Mr.
> Endo had was related to limitations of the DOM specification that only
> allows you to look up single elements by id.  One way to solve this
> and use the DISL scheme would be to use a different attribute as the
> name of the variable and have an id attribute that's sole purpose is
> to uniquely identify the element within the document.
>
> Whatever we decide, I believe we should choose a consistent scheme to
> handle this for <property>, <variable>, and <param>.  Thus using id in
> the declaration and name for the references may serve for this version
> of the specification with further improvements to come in the next
> version.
>
> Just my thoughts :)
> Jim