Shelby's Final Response to Tim Berners-Lee Regarding Semantics

As promised[3], I am responding to Tim Berners-Lee's post.

Note, I will not be answering any debate on this thread.  The following is
my final position (for near future) on the issue of semantics and presentation.

<h1>Shelby's Final Response to Tim Berners-Lee Regarding Semantics</h1>

<h2>Do TBL and I Agree?</h2>

<p>Tim Berners-Lee (TBL) posted[1] some opinions on semantics, as it
related to a discussion Ian Hickson and I were having regarding semantics
relative to XBL.  I assume we all know that TBL was instrumental in
creating the WWW and is noted expert on semantic web.  I am not clear if
TBL read the entire thread, leading up to my single post he cited.  I
assume it is possible that TBL did not fully read the context (whole
thread) within which I quoted from his web documents.  I am assuming he is
very busy and has limited time fo dwelve into all threads of W3C.  I also
assume he avoids "political" discussion.  Since I was unable to speak to
TBL to clarify my views, I will clarify here.  I think when my entire view
is taken into consideration, it is even more clear to me that I agree
factually with TBL's statements about semantics, both on his web documents
and in his post here[1].  And more importantly, that my positions against
merging semantic _controlling_ layers with presentation layers[2] is even
more consistent with TBL's position on the importance of consistent,
standard _implementation_ and _use_ of semantics as specified in social
contracts (e.g. laws or authoritative standard specifications).  My
position was never for anarchy.  To the contrary, I feel I am fighting
against anarchy and fighting for more relevant standards, with my efforts
to oppose the mayhem that might arise from inappropriate merging of layers.</p>

<h2>Why Did I Quote TBL?</h2>

<p>In discussing whether CSS was the correct layer to do XBL binding, Ian
Hickson and I got into a lengthy "discussion" to try to define what defines
and controls semantics.  I was asserting that CSS is presentation and that
presentation is not able to implement or change semantics.  I asserted that
XBL is able to implement and change semantics.  Since semantics is expected
to be consistent at the markup layer, then I asserted that any web
programming architecture capable of implementing (new tags) or changing
semantics, should be orthogonally in the markup layer (above DOM and
parsing of markup).</p>

<p>My understanding of Ian Hickson's position was that he felt nothing
could change or implement semantics, because he asserted that they are
under complete control of the specification for markup.  I responded that
at least for the case of new tags that might be implemented by XBL, then
the implementation itself might be the de facto social contract (the
specification) in force.  Ian responded that "new tags have no meaning".  I
think he may have later clarified that he meant new tags have no relevant
meaning to HTML specification, but that was well after I had quoted TBL.</p>

<p>Ian and I were also discussing one of Ian's examples, wherein Ian was
using XBL and regular expression parsing to interpret the content within an
"<a>content</a>" tag, as a date.  I asserted that he was by passing the
parser and markup layer and using scripting to essentially bind the markup
and presentation layer in "non-standard" ways.  In this context, I mean
"non-standard" because other layers will not have opportunity to understand
the semantics he was applying in his scripting.  I asserted that for the
semantic web, it was better he markup the date using a custom tag
"<mydate>" and use an XSLT transform to implement his special semantics of
that content.  I asserted that the <mydate> tag (but not the content? what
could be done about this?) would be ignored by UAs which did not understand
or implement the suggested XSLT layer.  Ian responded that his example was
purely presentation and that no semantics were being changed or defined by
parsing the sub-content as a date.  Realize that although I have suggest
XSLT as an alternative solution, that it may not be the ideal one, so
refuting XSLT _alone_ for this purpose, does not justify XBL.</p>

<p>I quoted[1] TBL to make the point that by allowing social contracts for
semantics of varying scope (internalized applications, small intranets, up
thru www and beyond) to happen in decentralized[4] manner, then it is
essential to success.  I quoted[1] TBL to provide a basis for new tags
having meaning on varying scopes.  This does not mean that those tags have
to interfere with tags that have W3C scope.  I also assert[2] that
transforming markup to markup preserves the W3C scope better than XBL
approach.  And that if Ian had a special semantics he wanted to add on top
of the HTML layer, then I asserted he should put it in the layer on top of
HTML, and not by creating a web programming architecture to change
semantics from the presentation layer.</p>

<h2>Does CSS Already Have Capability to Change Semantics?</h2>

<p>No[2]. Please read the cited reference carefully for the test I use to
answer that question.  Although Ian did point out one obscure example
wherein CSS can render an empty paragraph with an outlined box, and the
HTML spec says that empty paragraphs should not be rendered.  I think CSS
is still not rendering the paragraph.  It is rendering an empty box.  In
any case, one would need very compelling examples of CSS capabilities to
alter semantics, and in that case I would argue those capabilities would be
mistakes in CSS (if you agree with my test[2]), if they exist.  The whole
point of CSS was to be __optional hints__ to the presentation.  Note I got
ridiculed by Daniel Glazman for making that point[7].  Presentation should
not be involved in semantics that are specified for and by markup layer.</p>

<p>As far as I know, one of the main thrusts of CSS movement was to get the
presentation layer out of the markup layer.  Do we want to go backwards by
merging them again in complex ways?</p>

<h2>My Vision</h2>

<p>So why does keeping layers orthogonal outweigh the conveniences of
specialized architecture?  First of all, TBL's Principle of Least Power[5]
and his Principle of Modular Design[6] point out abstractly that keeping
layers well focused (least power for main goals) and modular (separated and
orthogonal) is essential to a scalable web design.  I quoted these in my
first post[9], and I was again ridiculed by Daniel Glazman in response[8].</p>

<p>More importantly, my vision is being able to quickly build a plethora of
new user agents (UAs) simply by mixing and matching (open source)
components.  The components should be small, well documented, efficient,
and orthogonal.  And I envision more plethora of layers built on top of
that with authored architecture.  I envision a very rich web.  Apparently,
Ian Hickson prefers a centralized web controlled by smaller (never greater
than number of web authors), backlogged (e.g. Mozilla) teams of UA
programmers[10].</p>

-Shelby Moore


[1] http://lists.w3.org/Archives/Public/www-style/2003Jan/0190.html
(follow TBL's citation of my quote and read multiple posts from Ian and me
around that time)

[2] http://lists.w3.org/Archives/Public/www-style/2003Jan/0147.html

[3] http://lists.w3.org/Archives/Public/www-style/2003Jan/0191.html

[4] http://www.w3.org/DesignIssues/Principles#Decentrali

[5] http://www.w3.org/DesignIssues/Principles#PLP

[6] http://www.w3.org/DesignIssues/Principles#Modular

[7] http://lists.w3.org/Archives/Public/www-style/2003Jan/0003.html

[8] http://lists.w3.org/Archives/Public/www-style/2002Dec/0188.html

[9] http://lists.w3.org/Archives/Public/www-style/2002Dec/0171.html

[10] http://lists.w3.org/Archives/Public/www-style/2003Jan/0198.html
(Ian wrote: "I'd rather have one team per UA implement ...")

Received on Thursday, 9 January 2003 05:10:01 UTC