W3C home > Mailing lists > Public > www-style@w3.org > August 2001

Re: CSS1 compliance/support (was Re: @version rule)

From: Tantek Celik <tantek@cs.stanford.edu>
Date: Thu, 02 Aug 2001 09:28:59 -0700
To: Ian Hickson <ian@hixie.ch>
CC: Bjoern Hoehrmann <derhoermi@gmx.net>, Jeffrey Yasskin <jyasskin@hotmail.com>, www-style@w3.org
Message-ID: <1215352810-369097378@psdbay.com>
From: Ian Hickson <ian@hixie.ch>
Subject: Re: CSS1 compliance/support (was Re: @version rule)
Date: Sun, Jul 29, 2001, 8:48 PM

> On Sun, 29 Jul 2001, Tantek Celik wrote:
>>>
>>> On Mon, 30 Jul 2001, Bjoern Hoehrmann wrote:
>>>>>
>>>>>     Which level should IE6 claim to support?
>>>>
>>>> Microsoft claims full CSS Level 1 compliance (or support, don't
>>>> know the exact wording).
>>>
>>> I said _should_. I know Microsoft claim to have full CSS1 support.
>>> I also know that unless their final release has significant
>>> improvements over their last preview release, they are not even
>>> close.
>>
>> I can only assume you are trolling Ian, so I'll bite. ;-)
>
> Doh, I did say that I only picked IE6 at random!

Ok, fair enough.

>>> Don't take this the wrong way, I'm very impressed by the
>>> improvements that IE 6 Preview has over IE 5.5 in terms of CSS1
>>> compliance. However, it is nowhere near the compliance of the best
>>> browser(s) on the market, and those browser(s) rightly do not claim
>>> full compliance.
>>
>> As measured by the W3C CSS1 Test Suite (the only official valid way
>> to measure compliance to CSS1 - unless you have a better
>> suggestion), IE6/Windows has full CSS Level 1 compliance,
>
> I am aware of 6 CSS1 bugs in the last WinIE6 Preview release that are
> visible in the CSS1 test suite.

I suppose I'll have to ask you to reassess your opinion when an updated
IE6/Windows release is available.

>> just as IE5/Mac did when it shipped over a year ago,
>
> I am aware of 1 CSS1 bug in the MacIE5 release that is visible in the
> CSS1 test suite.
>
>
>> which was certainly not true for other browsers at the time (among
>> others, the section 4.1.4 test [1] was particularly brutal, on even
>> those browsers who at the time were claiming "superior
>> conformance/compliance" - the results can be seen in Eric Meyer's
>> leader board).
>
> I am painfully aware of the many CSS1 bugs in Netscape 6.01 and
> Netscape 6.1 Preview 1 that are visible in the CSS1 test suite. (I
> can't give you an exact count since I don't track Mozilla's CSS1 test
> suite bugs any closer than CSS1 bugs found in other test suites. But I
> would guess there are between 1 and 6.)
>
> On the plus side, Netscape 6.1 [3] passes the 4.1.4 test [1] with only
> three bugs: there is very minor text overlap in the test after the 3rd
> long rule separator, the floats are not placed high enough when they
> are not at the immediate start of a line (as seen e.g. in the last
> test), and 'clear' isn't applied to <hr> elements in quirks mode (that
> last bug probably shouldn't be considered a bug, since it works in
> standards mode -- the test suite triggers Netscape 6.x's quirks mode).

See Eric Meyer's article on DocType switching in IE5/Mac - the test suite
DOES NOT trigger IE5/Mac's quirk's mode, as a test suite should be
considered strict.

> WinIE 6 Preview (6.0.2479.0006) has two bugs on this page: it doesn't
> put text high enough (e.g. in the third test <div>, the first one
> with the three floats),

While I agree with you that the CSS1 spec _allows_ for such an
implementation (putting the text up higher as you say), it does not require
it.  I have to double-check, but the language in CSS2 might be tightened up
a bit and require that, but I'm not certain.  In either case, that is not a
bug according to CSS1.

> and it doesn't put floats high enough (the
> same bug as seen in Netscape 6.1).
>
> MacIE5 (2022) also has two bugs on this page: First, it has the same
> text layout problem as WinIE6 on the third test <div>,

Which is not a problem according to CSS1 - only your interpretation of how
float would optimally work. (Note: I don't necessarily disagree with you
that the visual effect you describe is better - it is just not required by
the spec, one of several areas of the spec where implementations have an
opportunity to provide results better than that mandated by the spec.
:first-letter is another such area.)

> and second it
> puts unexpected spacing around the floats in that particular test.

I have seen this intermittently but have not been able to consistently
reproduce it - perhaps this is the reason Eric Meyer gave our float
implementation a (Q)uirky rating.

Nevertheless, I just tried loading that page (and resizing the window which
causes reformatting), and did not see any unexpected space.  I have seen
that in the past though.  If you find some set of conditions under which you
can cause the problem consistently, please let me know so what they are so
that we can correct it.

>> If other browsers pass the entire W3C CSS1 Test suite, then they too
>> could claim a similar level of conformance.
>
> Well. The CSS1 test suite doesn't test:
>
>    1. 'float' with 'width:auto'
>    2. the inline box model
>    3. the background/border issue mentioned recently on this list
>
> ...all of which should IMHO quite significantly impact a claim of
> "full CSS1 support".

Test suites are the only objective means we have to test/claim
conformance/support.  If you want to supply an alternative objective means,
then by all means, please propose it.  I have not seen one proposed thus
far.

> Like I said, I am not in any way rubishing the efforts made by
> Microsoft, or Opera, or Netscape, or any of the other browser makers
> out there.

Nor am I making light of the edge cases of CSS1 that you bring up - they
should be tested, but they are not yet in a test suite, so we have no way of
objectively discussing conformance/compliance to them.

> I just have Issues with anyone who claims "full support"
> merely on the basis of a basic test suite.

Like I said, provide an objective alternative then.

> I also have Issues with
> people claiming that they fully pass the CSS1 test suite when they
> don't, but I agree that it can be hard to be sure... ;-)

I think part of the issue in this case is your reading requirements into
CSS1 which are neither in the spec, nor stated in the test suite.

Specifically the descriptive text from the third DIV from sec414.htm[1]
says:

<blockquote>
This is text within the DIV. The floating elements within this DIV should
not go beyond the left or right inner edge of this DIV, which means that the
borders of the floating elements should not overlap the top or side borders
of the DIV in any way. In addition, they should not overlap each other. The
first two should be side by side, whereas the third should be on the next
"line."
</blockquote>

No where in that text does it say "how high is high enough or not" for the
text itself.

I recommend that you provide input to the CSS2 Test Suite so that it can
consider the additional requirements that you are requesting, as long as you
can demonstrate that those requirements exist in CSS2 (since they don't in
CSS1).  And the requirements you want don't exist in CSS2, then suggest them
for CSS3.

Tantek

[1] http://www.w3.org/Style/CSS/Test/current/sec414.htm

----------------------------------------------------------------------------
Dodge this.                                 http://www.microsoft.com/mac/ie/
Received on Thursday, 2 August 2001 12:28:46 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 27 April 2009 13:54:10 GMT