Re: <STYLE> considered harmful

Luke ~{B7?M~} (ylu@ccwf.cc.utexas.edu)
Sat, 18 Mar 1995 18:09:19 -0600 (CST)

On Sat, 18 Mar 1995, Brian Behlendorf wrote:
>On Sat, 18 Mar 1995, Luke Y. Lu wrote:
>>I consider this STYLE tag harmful and redundant.
>>
>>Reasons:
>>
>>1) It's illogical. Style is considered as attribute of content. This
>> tag treat it _as if_ it _is_ content.
>
>This is funny to hear coming from a fan of netscape tags :)

This is irrelevant -- again!. Besides, all the netscape tags implemented
so far are logical (yes, including BLINK -- an attribute of text), though
lack of hierarchy. Please refrain from dump irrelevant "personal feelings"
in later discussions. Thank you. :)

>There has been hum and cry from many people for wanting style information
>embedded directly in the document - this is an allowance for that. One could
>also say that the META tag and the TITLE tag aren't content, but rather an
>attribute of the document.

META has only attribute because it has no content i.e., it's defined as
<!ELEMENT META - O EMPTY>. Whatever stuff between <STYLE> and </STYLE> is
treated as content because it's defined as <!ELEMENT STYLE - O (#PCDATA)>
regardless of your interpretation. For the same reason, Stuff between
<TITLE> and </TITLE> _is_ content -- a content of highest hierarchy of a
document. The shortest document that convey a meaning is that with a
meaningful title. Stuff in STYLE can be 100K and tell you zero about
content.

>>2) Directly embedding non-html stuff in html is a bad idea. we'd
>> better refer them as foreign objects.
>
>I don't see it as "non-HTML" any more than I'd see someone putting C code
>between <PRE>'s into an HTML document as "non-HTML". What's between <STYLE>

You can't put C code between <PRE>, you need to convert <, > to &lt;, &gt;
respectively. You C code _is_ html after the conversion. And that's why
LISTING and XMP are depreciated.

>>3) <LINK rel=style href="style1.dsssl"> is more than enough for the
>> purpose. we can use more than one link to merge styles. With local
>> caching, unnecessary merge operation can be avoided.
>
>As Lou from netscape and others have pointed out, this is a big network
^^^^^^^^^^^
>problem if every document with style information were forced to make an extra
^^^^^^^

I bet you add the "big" there. It's no worse than an inline image.
Besides, by the time style is prevalent, httpd servers that support
multiple part MIME type get per request will be prevalent. And it
will cease to be a "network problem".

>>4) Incompatibility. clients that do not recognize this tag will display all
>> the dsssl-lite garbage on the screen because it is treated as if it's
>> document content (see reason 1)
>
>Nope, it's in the <HEAD>, so even a conformant 2.0 browser won't display it
>at all. Netscape and I believe all the Mosaics handle this fine.

Yeah, you "believe". Have you really tried? lynx, netscape all displayed
the dsssl-lite garbage. Besides, you know most html pages out there
conveniently omit HEAD...

__Luke

--
Luke Y. Lu
mailto:ylu@mail.utexas.edu/
http://www.utexas.edu/~lyl/