RE: STYLE_SHEETS_USE test

Hi Jo,

Hi Jo,

Some comments inline:

>So the most we can and should do is warn -- unless there is a basic 
>syntax error like

>zillon {planet:tharg {

>in which case we should FAIL.

>Can you tell the difference between that type of error and one of the 
>STYLE_SHEET_USE types of error?

Yes, I could distinguish (CSS Validator does) between parsing errors and
the other kind of errors that STYLE_SHEET_USE defines (these are raised
as warnings).

The problem is that, AFAIK, CSS Validator *does not* distinguish between
the subtests of STYLE_SHEET_USE. I mean, if we consider the example you
put in your last mail:

>There is a difference between STYLE_SHEET_USE-5, STYLE_SHEET_USE-6
>color: orangey-blue;  /* not a recognized value */
>color: thin; /* a recognized value but not appropriate to color */

CSS Validator would provide us the following output:

<warning>
       <info>orangey-blue is not a color value</info>
	[...]
</warning>
<warning>
      <info>thin is not a color value</info>
        [... ]
</warning>

So this is our main concern about this test. We would have to implement
an ad-hoc solution that would wrap the output of CSSValidator and then
we should treat it in order to distinguish each subtest case. 

Unless I am missing something, this would be a considerable development
effort taking into account the available resources (time and human).

Abel.


-----Mensaje original-----
De: Jo Rabin [mailto:jrabin@mtld.mobi] 
Enviado el: viernes, 04 de julio de 2008 16:35
Para: Abel Rionda
CC: public-mobileok-checker; public-bpwg-comments@w3.org
Asunto: Re: STYLE_SHEETS_USE test

Hi Abel

Aaargh, now that you point it out, the text of this test needs changing.

Specifically the Note is wrong as STYLE_SHEET_USE-5 specifically tests 
for properties that are not defined in CSS Level 1. So delete that note.

There is a difference between STYLE_SHEET_USE-5, STYLE_SHEET_USE-6

color: orangey-blue;  /* not a recognized value */

color: thin; /* a recognized value but not appropriate to color */

color: 22; /* ditto */

If anything I'd say that STYLE_SHEET_USE-7 and STYLE_SHEET_USE-8 are 
both contained in STYLE_SHEET_USE-6.

border-width: 22; /* requires a unit but doesn't have one */

border-width: 22Hz;  /* has the wrong type of unit, though actually I am

not sure this can happen in CSS 1 */

I would not be at all bothered if we combined all these into a single 
clause, however it would have to remain a warn ...

... ref the checker's CSS validator. The whole discussion about CSS (if 
I recall correctly) came up as a Last Call comment from Bert Bos. And 
(again, if I remember correctly) we agreed to change. I don't know how 
we could ignore that now. In any case, it is quite simply wrong to fail 
on unknown properties etc. it's very much the intention of CSS that a 
CSS2 style sheet works just fine in CSS1 simply by ignoring the stuff 
that is not understood.

So the most we can and should do is warn -- unless there is a basic 
syntax error like

zillon {planet:tharg {

in which case we should FAIL.

Can you tell the difference between that type of error and one of the 
STYLE_SHEET_USE types of error?


Jo

On 04/07/2008 12:25, Abel Rionda wrote:
> Hi Jo,
> 
>> This is particularly fraught since CSS is by design open-ended so 
>> outside of the syntax requiring certain combinations of valid tokens
> and 
>> punctuation, almost Anything Goes [1].
> 
> Yes, indeed. BTW, great song that reminds me of the opening scene of
> Indiana Jones and the Temple of Doom...
> 
> So the definition of valid CSS under 2.4.8 is a simple syntax checking
> without lexical checking at all. I think that call this "CSS Valid" is
a
> bit confusing for the user (CSS Valid would be inferred by the user as
> *valid* as the spec of the CSS Level x states). 
> 
> Besides the special CSS Validity, STYLE_SHEETS_USE is still
duplicating
> some checks.  This test has a note regarding CSS properties that are
> defined or not on Level 1:
> 
> "Note: The tests on CSS property values only apply to properties
defined
> by CSS Level 1; other properties are ignored for the purposes of this
> test."
> 
> Having this note on mind and applying the following checks, 
> 
> --
> "If the CSS Style contains at-rules (other than the @media at-rule,
and
> the media list of the @import at-rule), properties, or values that are
> not recognized as being specified in CSS Level 1, warn
> 
> If the CSS Style contains a property with a value that is
inappropriate
> to it, warn"
> --
> Some questions come to my mind:
> 
> Is the same "not recognized" and "inappropriate"?, could a "not
> recognized" value be appropriate?, I don't think so.
> 
> The first check contains the second one. Any warn reported by the
second
> will be also reported by the first. There are the following
> possibilities:
> 
> - In case of a CSS Level 1 property with a "not recognized" value,
both
> checks apply.
> - If the case is a non CSS Level 1 property the first check will
report
> a warn but, as stated by the special note, the second one will be
> ignored.
> 
> 
> Also both "not recognized" and "inappropriate" terms should be
> clarified, i.e. is a font-size value of 1000px inappropriate? , is
> "inappropriate" used as a synonym of a non CSS Level 1 valid value?
> 
> 
> Anyway, I think we are going to run into problems in the current
checker
> implementation:
> 
> -I do no see a way to deactivate these features in our CSS validation
> tool (CSS Validator makes a complete grammar validation) and making an
> ad-hoc solution would tale a considerable development effort.
> Particularly, STYLE_SHEETS_USE test (currently these subtests are not
> made) is going to be hard to implement basing on the checker design
and
> our limited time to dedicate to this.
> 
> I am not sure what the best decision on this is if we want to release
> checker in a reasonable time.
> 
> 
> Regards,
> 
> Abel.
> 
> 
> 
> -----Mensaje original-----
> De: Jo Rabin [mailto:jrabin@mtld.mobi] 
> Enviado el: viernes, 04 de julio de 2008 11:00
> Para: Abel Rionda
> CC: public-mobileok-checker; public-bpwg-comments@w3.org
> Asunto: Re: STYLE_SHEETS_USE test
> 
> Hi Abel
> 
> We jumped through a couple of hoops of fire around this subject
possibly
> 
> during LC-1 or LC-2 (the mobileOK Jurassic Period, perhaps). As far as
I
> 
> can recall the question was around exactly what is meant by "valid 
> CSS1". This is particularly fraught since CSS is by design open-ended
so
> 
> outside of the syntax requiring certain combinations of valid tokens
and
> 
> punctuation, almost Anything Goes [1].
> 
> Consequently, we have the definition of valid CSS under 2.4.8:
> 
>      A resource is considered a valid CSS resource if it conforms to
the
> 
> grammar defined in [CSS], Appendix B (see note below), except that 
> @media at-rules, which are not part of the grammar, are allowed and
are 
> not considered invalid. The presence of at-rules, properties or values

> or combinations of properties and values that are not specified in
[CSS]
> 
> does not constitute a validity failure for CSS. See 3.21 
> STYLE_SHEETS_USE for treatment of such values. In addition, the @media

> at-rule and the presentation media list for the @import at-rule are 
> taken into account when evaluating CSS.
> 
> [I've just  spotted that the clause "see note below" is a dangling 
> reference and needs to be removed.]
> 
> So I think what is happening here is that the checker CSS Validation 
> that is being carried out is actually stricter than that implied by 
> CONTENT_FORMAT_SUPPORT and is potentially mis-reporting inappropriate 
> combinations of properties and values as failures under that heading. 
> They should in fact be reported under STYLE_SHEETS_USE.
> 
> Note particularly that it is not invalid to use properties that are
not 
> known in CSS1, e.g.
> 
> klingon {foo: bar;
>      distance: 3light-years;
> }
> 
> is valid, though it contains properties values and units that are not 
> defined in CSS1.
> 
> Hence the warns rather than the failures in STYLE_SHEETS_USE.
> 
> Jo
> 
> 
> [1] http://en.wikipedia.org/wiki/Anything_Goes_%28song%29
> 
> 
> 
> On 04/07/2008 09:17, Abel Rionda wrote:
>> Hi,
>>
>> While we were reviewing the test implementation status in checker
> code,
>> we found out some checks 
>> of STYLE_SHEETS_USE regarding CSS values that we would like to
comment
>> [1]:
>>
>> [begin STYLE_SHEETS_USE fragment]
>>
>> If the CSS Style contains a property with a value that is
> inappropriate
>> to it, warn
>> If the CSS Style contains a property with a value that requires a
unit
>> or a percentage:
>>       If the unit (or percentage) is not present, warn
>>       If the unit (or percentage) is inappropriate to the value, warn
>>
>> [end STYLE_SHEETS_USE fragment]
>>
>> All these checks are already made during grammar validation test
>> (CONTENT_FORMAT_SUPPORT) and they would raise a *FAIL* (while in
>> STYLE_SHEETS_USE at most we would get a *warn*)
>> We do not see any benefit of this duplicity and, furthermore, due to
>> they raise different level errors, it might lead the user to
> confusion.
>>
>> Regards,
>>
>> Abel.
>>
>> [1]
>>
>
http://www.w3.org/TR/2008/WD-mobileOK-basic10-tests-20080610/#STYLE_SHEE
>> TS_USE
>>

Received on Friday, 4 July 2008 16:36:06 UTC