Re: [1.2T-LC] 16.2.9 values attribute 'extended syntax' (ACTION-2318, ISSUE-2084)

Hi, Dr. Olaf-

The SVG WG reopened this issue, and based on the weight of your
argument, agreed that the short-term fix for the funky content was not
worth the potential pain of future incompatibility.  Ikivo has verified
that the next release of their Animator authoring tool will fix this
bug, as well.

Therefore, we have removed the paragraph about the trailing semicolon
for the 'values' attribute, as you've requested:

 http://dev.w3.org/SVG/profiles/1.2T/publish/animate.html#ValuesAttribute

Please let us know if this resolves your comment in a more satisfactory
way.  We appreciate your insight and diligence on this matter.

Regards-
-Doug


Dr. Olaf Hoffmann wrote (on 10/19/08 11:47 AM):
> Hello Doug,
> 
> up to yesterday I had only the impression, that this new 
> 'extended values syntax' with additional superfluous semicola
> is 'just' incompatible with SMIL and causes nasty problems with
> valid empty values.
> After some thinking I started yesterday to write some tests to
> see, what may happen, if this is really implemented and my
> current impression is, that this is not really implementable 
> including all side effects on SMIL animation and maybe I even
> did not find all side effects (I currently tested only simple ones)
> ....
> 
> Therefore, if this remains in the animation chapter, I think
> some advanced tests are required to see, if it is really (not)
> implementable. I think, for such a new feature at least two
> implementations are required.
> Without any surprise the animation capable programs,
> I have installed, do not survive any of the tests I created 
> (Adobe plugin, Opera, Batik/Squiggle 1.7, ...).
> 
> 
> 
> Doug:
>>
>> Actually, the situation is that one of the more popular SVG animation
>> authoring tools, Ikivo Animator, makes this mistake in its output.  This
>> resulted in a large amount of faulty content.  Because this was a fairly
>> obscure point, SVG 1.2 Tiny viewers adapted to the content.  
> 
> Really? I have serious doubts about this.
> Because I have no program, which implemented this somehow
> completely or at all, I would like to know, which programs survive
> really basic tests about this new feature (see attachments).
> 
> Maybe if the Iviko Animator itself has some display and not just an
> (erratical) output of source code, this would be the first suspect
> to check, if it  'eats' his own products as described in the current draft ;o)
> 
> If there is another (mobile) viewer having implemented this as described,
> this should be tested and reported too ;o)
> 
> 
> 
>> However, we 
>> have spoken to Ikivo about this problem, and they have agreed to change
>> this error in future releases of their product.
>>
> 
> This sounds fine.
> 
>> In the meantime, we have to accommodate the current behavior of viewers,
>> and to inform authors about the potential pitfall of the trailing
>> semicolon syntax.  I have therefore changed the passage in question to
>> permit, but not mandate, this behavior, but also to caution authors and
>> authoring tools against using this syntax.
>>
>>  http://dev.w3.org/SVG/profiles/1.2T/publish/animate.html#ValuesAttribute
>>
>> We will examine this more deliberately in SVG 2.0 Core.
>>
>> Please let us know if this satisfies your comment.
>>
> 
> It would be no problem, if it would only affect previously invalid syntax in 
> attribute values, however because it is possible to have empty list items
> too (as indicated too in the draft), this has influence on valid content 
> too and corrupts somehow the propper interpretation of valid content.
> 
> SMIL animation is pretty complex and is mainly based on the values
> syntax, therefore there can be even more side effects on valid content,
> for example on the equivalence principle of from-to, from-by and by animations
> with the related values animation.
> 
> What about invalid syntax for other lists? 
> Does the Ikivo Animator generate propper lists for keyTimes, keySplines,
> keyPoints or does it have the same problem for those lists? 
> Note, that currently it is for example specified, that those lists are
> currently defined to be ignored if they contain nonsense like the
> superfluous semicolon. 
> I think, even this was never tested in any test, but gets more relevant
> due to this 'extended values syntax' and that such an error handling is
> propper implemented for those lists, it important to avoid more faulty
> content. The rules for these lists are changed too compared with 
> SVG1.1 and in a different way than now for values. And if some
> implementation already implemented this 'new' values syntax before it
> was specified, maybe they ignored these rules for the other lists too? 
> Because all these things are not indicated to be errors anymore, they
> turn into features and should be tested to convince everybody, that
> such features are really implementable.
> 
> I attached three simple tests related to this and it would ease my
> serious doubts about the implementability of such new syntax, if
> at least some viewers are reported to survive those tests and
> maybe some more related tests concerning more possible side
> effects of this 'extended values syntax'.
> Maybe others have more ideas, what could go wrong with this
> syntax and provide tests about other side effects.
> 
> Assuming, that this is implementable, the new note, that
> 'authors and content generation tools are discouraged from 
> using trailing semicolons' might be sufficient for SVGT1.2 to
> avoid more nasty content and incompatibilities of this type,
> however it is a little bit surprising, that implementors want to
> implement this complex temporary handling of a really stupid
> error and have problems to agree with fixes of other really 
> nasty things ;o)
> 
> Olaf
> 
>

Received on Tuesday, 4 November 2008 13:32:18 UTC