Not to be overly picky (because I agree with the general point), but
scripts
of the form outlined below actually can be transformed in a general
way
without the need to perform flow analysis or know anything about
the
semantics of the script itself.
The ones that are more bothersome are scripts of the form:
<script>
document.write("<font
size='6' color='green'>Portlet1</font>");
</script>
If, as a Consumer, I want to customize the size and/or color of the
font,
I'm out of luck without some help (as you point out). Of course,
even if
this is given to me in nice, clean markup, I don't really know the
significance
(at the application level) of the <FONT> element, so I still need
something
that tells me that this is the FONT used to display part numbers or
sale
prices or ....
To me, it seems like a matter of degrees, not true difference.
Which is
why I hesitate to support the additional wording. It takes pains to
draw a
distinction between script and markup which, although it supports a
common perception, isn't really there. Also, I think the emphasis
on action
routing is misplaced (don't get me wrong, that's important too, but so is
general customization).
As to the potential solution list, I think I agree with the
choices. Option 1
is not an option in this case (the capability is much too core).
Options 2
and 3 come back to the age old "who does the work"
question. To me the
only correct answer is "yes", since both approaches have pluses
and minuses
and I believe that we will want to support distribution of this work
throughout
the network (from Producer to Consumer to intermediaries). I don't
think we
want this requirement to favor one topology over another (and I don't
think
it does).
Sean
At 06:38 PM 5/8/2002 -0400, Eilon Reshef wrote:
My
thought was around people that use scripts like this:
<script>
var target =
"http://" + gMyDomain +
"/mypage.rgb?page=" +
document.form[1].pageNumber.value;
document.location
= targert;
</script>
My intent was to
formulate a part of the requirement that clearly states that WSIA cannot
expect Consumers to automatically rewrite this action to point back to
the Consumer.
Does this make sense as
the intent of this part of the requirement?
As for a proposed
solution, one can consider three options:
1. Disallow such cases
and other such constructs (which is always the last resort).
2. Define a
meta-language that does not require changes to the script (assuming it's
a script that was written beforehand), and tries to capture the
replacement locations externally (a-la adaptation). However, the halting
problem (;-) does argue that there will still be cases that the external
language won't be able to address.
3. Ensure that enough
information is passed to the Producer so that when it either writes new
applications or when it needs to adapt cases like this, it would be
technically possible.
Any
thoughts?
Eilon
-----Original Message-----
From: Sean Fitts
[mailto:]
Sent: Wednesday, May 08, 2002 5:57 PM
To: Eilon Reshef;
Subject: RE: [wsia][wsia-requirements][R602]
At 04:51 PM 5/8/2002 -0400, Eilon Reshef wrote:
I think I see your point (I thought of modification as semantic
versus syntactic), how about the following wording (which still puts the
Customization issue aside):
One small tweak, the modifications may be semantic, but the semantics
are
largely implied and will likely need to be described externally (ala
the adaptation
description in WSXL).
I guess I don't see a lot of difference here between this type of
modification and
general markup modification. Both contain semantic information,
both require
some sort of locator to identify the sections implementing a given
semantic
operation, and in both cases, the semantics are opaque.
This specification must support common Presentation formats, which
are in use today in Net-enabled applications. In particular:
1. It MUST support Presentation Fragments in HTML, XHTML, XML and
WML.
2. It MUST support JavaScript as an associated scripting language.
Such support MUST include a way to support Actions triggered by scripts.
However, it SHOULD NOT be assumed that the Consumer is aware of the
semantics of scripting elements.
Given the parallel between semantics of scripting elements and
semantics
of other elements (such as markup and/or actions), I still don't see
the need
for the second statement. However, in this form it seems a lot
more benign
(at least to my eye).
3. It SHOULD support embedded binary presentation elements (e.g.,
Flash, Applets, etc.).
[Optional/Debate: Such support SHOULD provide a way to support
Actions triggered by such elements.]
However, it SHOULD NOT be assumed that the Consumer modifies the
binary elements in any way.
I personally think it makes sense to favor a single technical
approach that captures both (2) and (3), but I also don't see it as a
high-level requirement but rather as a technical
preference.
Could you elaborate on this? I'm not sure I understand.
Thanks.
Sean
-----Original Message-----
From: Sean Fitts
[mailto:]
Sent: Wednesday, May 08, 2002 2:14 PM
To: Eilon Reshef; 'Rich Thompson';
Subject: RE: [wsia][wsia-requirements][R602]
At 01:55 PM 5/8/2002 -0400, Eilon Reshef wrote:
I am not suggesting that the Consumer is not allowed to change
JavaScript, rather the suggestion is that we wouldn't assume that it
should. To me, that's because correctly analyzing code constructs (in any
language) without executing them is anywhere from hard (from a practical
perspective) to impossible (from a theoretical perspective, as Theory of
Computation shows).
I don't see a connection between supporting modification of
JavaScript
(which I agree is an open issue) and the need to support complete,
path wise analysis of it. Leaving the halting problem aside for
a bit, it
would seem possible to extend the Adaptation Description Language
proposed by IBM to include JavaScript modifications along with XML
and CSS ones.
This is not to say that WSIA can't define an interface that uses
JavaScript (e.g., I assume the committee may decide to define JavaScript
functions, events, etc.), but I guess that the question is can we require
the Consumer to analyze JavaScript code to support action routing, for
example?
Again, I don't see how leaving the second sentence out leads to
*requiring* the Consume to analyze or even modify JavaScript.
Such a
statement would seem to need a positive assertion that such
modification
*is* a requirement (something which again, I view as open).
Customization is definitely something that we will be discussing in
the Customization sub-committee. My working assumption is that the
requirement below is rather generic, and applies to anywhere from the
scope of WSIA in general, to action routing, unique tokens, etc., and
that it might be changed as the Customization sub-committee
proceeds.
I guess my take is that it is too generic. It seems to be
trying to
take a half step and would result in muddying things instead of
making
them clearer. If it really doesn't place any restrictions one
way or the
other on our work, then it doesn't seem like a requirement and I
would
argue it should not be included.
Sean
Eilon
-----Original Message-----
From: Sean Fitts
[mailto:]
2. It MUST support JavaScript as an associated scripting language and
MUST provide a way to support actions triggered by scripts.
[Optional/Debate: However, it MUST NOT be assumed that scripting elements
are modified by the Consumer in any way.]
Why do you feel that the second "MUST NOT"
statement is necessary?
To me it seems overly restrictive since it impacts both what types of
customization we will support and where the customization will occur.
My understanding is that both of these issues are still up for debate/
description in the customization sub-group.
