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

CSS DOM woes

From: Bjoern Hoehrmann <derhoermi@gmx.net>
Date: Sun, 30 Sep 2001 05:13:31 +0200
To: www-style@w3.org
Message-ID: <qb3drts04k70tqv5ffidcucftt2f9hm4og@4ax.com>
Hi,

   The more I try to implement the current DOM Level 2 Style
recommendation in Perl, the more I dislike it. I asked a lot of
questions in mid April 2001, but I didn't get any answer, so I ask them
again and new ones, this time bugging www-dom, www-style *and* the
editors (with exception of Vidur Apparao, the recommendation doesn't
list any of the editors email addresses and I don't know his... :-( ).

Genericalism:

  The CSS DOM is very focussed on CSS Level 2 thus making it hard or
  nearly impossible to implement it in a generic fashion. This is
  allowed at least implicitly by the CSSStyleDeclaration definition in
  the specification. So, my questions assume, that the implementation
  doesn't know anything about things specific to a certain level of CSS.

The cascade:

  Some methods provide access to properties by their name, this implies
  that the cascade has to apply somewhere, but it isn't specified where.
  What rules apply to a CSS DOM implementation? Is it

    * only the last declaration counts
    * !important declarations override non-!important declarations?

  If this is the case, a call like

    CSSStyleDeclaration::setProperty('coolness', 'root', '')

  doesn't do anything if there was already a
  'coolness: kiddie !important'?

Typing:

  What constitutes a CSSValueList? It can't be the property's definition
  (since generic implementations don't know nothing about them), so this
  information must lie in the CSS tokens, but how is this supposed to
  work in a case like

    font-family: Arial Unicode MS, Trebuchet MS, sans-serif

  We have 6 CSS_IDENT values here, there are no other CSS types, that
  could be used to represent the comma but CSS_UNKNOWN, however, the
  comma is not specified as CSS value in the normative reference CSS
  Level 2. The comma will be lost in the DOM, an application is unable
  to know about what values make a singe font family name, the cssText
  attribute is useless, since the DOM application is supposed to
  re-stringify the property, in fact, the original text isn't available
  if the DOM builder used SAC to parse the style sheet. The
  specification could have said commas to create new CSSValueLists, but
  CSSValueList::item returns only CSSValue object, so this is no
  possible implementation.

Constructors:

  Where are create* methods? Why is there no interface to create
  CSSValues, CSSValueLists, etc. Actually this question should be,
  where are the objects in this object model? I/O via strings may be
  ok for values and selectors, but certainly not for rules and
  declarations, this just complicates implementations. Stringifycation
  is a job for serializers, but not for a DOM.

Error Treatment:

  It is not specified, how implementations are supposed to treat errors
  in string input, raising a DOMException with SYNTAX_ERR is not
  compatible with the forwared compatibility rules of the normative
  reference CSS Level 2. Consider some application tries to set
  CSSStyleDeclaration.cssText to

    "font-size: 2em; font-family: 'sans-serif'"

  A CSS parser will ignore the infalid font-family declaration, the DOM
  Interface would probably raise the mentioned DOMException and ignore
  the whole declaration. Or would it set font-size? But how should the 
  application then catch the error?

Selector API:

  Why is there no API for W3C Selectors? How should applications find
  out about what rules they want to manipulate, if they can't find out
  if the selector matches their needs, or how are simple modifications
  to some selectors to be done? The XPath DOM3 module also lacks of a
  feature to access XPath expressions. I think some task force should
  try to give them both a shareable API; this should be possible, since
  they share functionality and have similar syntax. Of course they need
  also specialised APIs to access their full potential, but if they'd
  have something in common, the learning curve of both and
  implementations of possible translators between W3C selectors and
  XPath expressions could be built more easily.

CSSStyleDeclaration:
 
  attribute item: "Used to retrieve the properties that have been 
  explicitly set in this declaration block. The order of the properties
  retrieved using this method does not have to be the order in which
  they were set." How does this interact with setProperty()?  If I add a
  new property via setProperty(), which index number does it get? length
  i.e. it is inserted at the end of the 'list'? Or doesn't that matter,
  i.e. the order may be changed through calling setProperty() or
  removeProperty()? In general, it doesn't make much sense to make
  CSSStyleDeclaration some type of list if this list is only accessed by
  property names, but this is a general DOM disease.

Font Descriptors and more typing:

  Given ist the following style sheet:

    @font-face {
      font-style: normal, italic;
    }

  How to present this in the DOM? I _very_ dislike the current
  situation, that font descriptors are treated as properties, however,
  according to the CSS grammar we have here [IDENT, DELIM, IDENT] in the
  property value. In the DOM is a CSSFontFaceRule for the 
  @font-face-rule created and CSSStyleDeclaration for it's content. Ok,
  but what about the value? I thought individual tokens make individual
  CSSValues, then we have a CSSValueList with

    [CSS_PRIMITIVE_VALUE, CSS_UNKNOWN, CSS_PRIMITIVE_VALUE]

  Is that the way to go, or is it

    [CSS_PRIMITIVE_VALUE, CSS_PRIMITIVE_VALUE]

  only for the 'normal' and 'italic' IDENTs? Which DELIM tokens would
  then not go into the DOM? How to present something like

    elem { twinkle: 6 > 7 }

Value Unit Conversion:

  The units [cm, mm, in, pt, pc], [hz, khz], [s, ms], [deg, rad, grad]
  allow mutual conversion, are other conversions possible? Can a CSS_URI
  be a CSS_STRING? Can CSS_DIMENSION and CSS_NUMBER be converted into
  something else?

Unicode Ranges:

  The specification does not provide any way to access Unicode Ranges.
  Why this omission? Representing them as CSS_UNKNOWN types doesn't
  really make sense.

Actual Values:

  The current CSS DOM doesn't provide any means to access the actual
  value for properties, why?

I'm sure I've missed a lot of issues (maybe others post ), but I'd be
very happy to get some clarifications on the mentioned points. The CSS
DOM will become more important since SVG users are likely to use it, but
that's IMO impossible, with a CSS DOM specification like the one
provided.

regards,
-- 
Björn Höhrmann { mailto:bjoern@hoehrmann.de } http://www.bjoernsworld.de
am Badedeich 7 } Telefon: +49(0)4667/981028 { http://bjoern.hoehrmann.de
25899 Dagebüll { PGP Pub. KeyID: 0xA4357E78 } http://www.learn.to/quote/
Received on Saturday, 29 September 2001 23:14:39 GMT

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