RE: [SVGMobile12] Question on details of when <script> elements execute

Boris,

Below are answers to your questions posed in 
http://lists.w3.org/Archives/Public/www-svg/2006Jul/0046.html. 
I have not pruned the discussion. So I have prefixed my replies 
with [Andy] to make them easier to search for.

I believe this also answers all the issues to which you refer
in http://lists.w3.org/Archives/Public/www-svg/2006Jul/0107.html.

Please review this email and the updated WD and let us know
if they address your concerns as soon as possible.

Regards,
Andy

PS. Incidentally the comment was previously closed as I had 
thought I sent this Wednesday, but it got caught in my system. 


> -----Original Message-----
> From: Boris Zbarsky [mailto:bzbarsky@mit.edu] 
> Sent: den 17 juli 2006 07:18
> To: Andrew Sledd
> Cc: www-svg@w3.org
> Subject: Re: [SVGMobile12] Question on details of when 
> <script> elements execute
> 
> Andrew Sledd wrote:
> > Some factors that are involved:
> 
> I understand the issues, yes.  Having reread the reply to my 
> load event question, my remaining concerns are as follows 
> (the numbered items):
> 
> 1)  The load event definition says, "In the case the element 
> references a script, the event will be raised only after an 
> attempt to interpret the script has been made."  At the same 
> time, the script element definition says, "Script execution 
> happens just after the load event occurs for the given script 
> element."  These are contradictory.  Which one is correct?
> 
[Andy]  The load event description is correct. The script execution
description has also been updated to reflect this.

> > b) The load event associated with an element that references an 
> > external asset via xlink:href does not occur until the 
> asset is fully 
> > pulled down and available
> 
> 2)  What happens if the xlink:href is changed before the load 
> event fires? 
[Andy] only the original scripts content is executed as per the
spec. The change to xlink:href will raise an additional load
event when the external resource is loaded.

> Is the first load aborted and a new one 
> started?
[Andy] No. Both load events occur. Note however that only the 
original script content is executed. 


> > therefore, if scripting logic changes the <script> element, in some 
> > cases these changes will happen before the original 
> scripting logic is 
> > executed (in which case, the UA must only execute the newer 
> scripting 
> > logic based on the changed DOM)
> 
> 3)  This is what was not obvious to me from the 
> specification.  I believe the specification would benefit 
> from an explicit statement to this effect.
> 
[Andy] The key specification text is that a <script> element is
 "executed at most once"!

> > As a result, the advice to content creators: don't do that.
> 
> For what it's worth, that's an authoring guideline, whereas I 
> was looking for normative text from an implementor's perspective....
> 
> > Scripts should not change <script> tags because this might 
> result in 
> > unpredictable results.
> 
> 4) Only due to unpredictable interleaving with async events.  
> Given an event sequence, I feel that there should be a 
> predictable outcome and that the specification should specify it.
> 
[Andy] Most, and in the case the content is carefully constructed,
I believe nearly all, unpredictability is removed if 
externalResourcesRequired is set to true. In which case the content 
provider receives feedback thru the load event that the 
script-content has executed.

> > 2)  Do scripts created via createElementNS() ever execute?  If so,
> >      when?  Note that it's not clear to me whether the load event
> >      fires on nodes created via DOM methods.
> > 
> > Nodes do not become part of the SVG document until they have been 
> > inserted into the DOM tree.
> 
> Not really important, since I can go ahead and insert them, 
> but I think your reply about load events addresses the 
> concern here, modulo my remaining concern #1.
> 
> > 3)  Do scripts that are clones (via cloneNode()) of other scripts
> >      execute?  If so, when?  Does it matter whether the script that
> >      was cloned had already executed at the time of cloning?
> > 
> > Same answer as above.
> 
> 5)  Again, assume that the <script> is then inserted in the 
> document and let's reask the original question.  Note that 
> this is not a trivial question, since for HTML the de-facto 
> standard behavior is actually different for <script> nodes 
> created via cloneNode() vs those created via createElement().
>  I'm not sure whether the SVG WG wants to specify something 
> like that, but it may be worth thinking about, at least.
> 
> > The new wording is:
> > "Removing or altering script elements after script 
> execution has taken 
> > place has no effect. Inserting a script element into the 
> DOM tree will 
> > cause it to execute."
> 
> 6)  Still subject to the "at most once" constraint, right?  
> Perhaps worth saying this as:
> 
>    "... will cause it to execute if it has never executed before."
> 
> or something.

[Andy] This is meant to answer both 5 and 6; and I hope you agree.
A script element can be cloned or created and inserted.
Each time it is inserted it is executed. The clause "Inserting a 
script element into the DOM tree will cause it to execute." is not 
limited to new or unexecuted script-content. 

> 
> Looking forward to changes to resolve contradiction 1, 
> clarification wording for 2, 3, 4, a decision on 5, and maybe 
> clarification wording on 6.  ;)
> 
> Thanks,
> -Boris
>

Received on Monday, 24 July 2006 10:33:55 UTC