- From: Norman Walsh <ndw@nwalsh.com>
- Date: Wed, 12 Jan 2011 13:59:08 -0800
- To: public-html-xml@w3.org
- Message-ID: <m2lj2prdg3.fsf@nwalsh.com>
[ Thank you, Noah, for scribing. ]
See http://www.w3.org/2010/html-xml/2011/01/11-minutes.html
[1]W3C
- DRAFT -
HTML/XML Task Force
Meeting 3, 11 Jan 2011
[2]Agenda
See also: [3]IRC log
Attendees
Present
Norm, John, Yves, Michael Champion, Michael Kay, Noah, Henri
Regrets
James, Anne
Chair
Norm Walsh
Scribe
Noah Mendelsohn, NM
Contents
* [4]Topics
1. [5]Administrivia
2. [6]Use case #3, islands of HTML5-marked prose
3. [7]Use case #4, HTML document with islands of XML
* [8]Summary of Action Items
--------------------------------------------------------------------------
Administrivia
NW: Next call will be in a week, on 18 January. Any regrets?
Silence.
topics: Use cases
<hsivonen> Use case email was
[9]http://lists.w3.org/Archives/Public/public-html-xml/2010Dec/0064.html
NW: I've been somewhat out of touch, but have seen at least two
interesting email threads: 1) xml in feeds and 2) how to detect html5
JC: XML or HTML?
NW: Well, some thread subjects said XML
<hsivonen> we covered use cases 1 and 2. We didn't cover 3 and 4
Use case email was
[10]http://lists.w3.org/Archives/Public/public-html-xml/2010Dec/0064.html
Use case #3, islands of HTML5-marked prose
From the email description of the use case:
3. I have an XML document and I want to embed islands of human prose
marked up with HTML5 in it because I want to be able to extract
those sections for use in, for example, documentation.
JC: In that environment, we don't have an HTML5 DOM, I think, so we don't
have to deal with inconsistent DOMs
NW: Yes, mainly XML tools for this case.
JC: What limitations are there on HTML5? E.g., I know about noscript.
NW: (missed something about semantics) I was thinking about things like
HTML5 rules that automatically add namespaces to SVG, and that won't
happen in an XML toolchain.
JC: The XHTML5 elements mean the same as their like-named counterparts in
HTML5, with the exception of NOSCRIPT
HS: Yes, and also ISINDEX
NW: Why?
HS: Those are both sort of parser-managed things on the HTML side. ISINDEX
as sort of a parser macro, is invalid into HTML5, and is invalid in that
sense. It expands into other elements like a macro. NOSCRIPT depends on
the context.
JC: How it's parsed depends on whether you have scripting.
NW: Thanks, good to know. Sounds like it's safe to set aside ISINDEX. Less
sure about NOSCRIPT, but likely at worst a minor problem.
Use case #4, HTML document with islands of XML
From the use case email
4. I have an HTML5 document and I want to embed islands of XML in it
because I want to be able to write JavaScript and CSS to manipulate
those elements, for example, in the browser.
NW: The HTML5 parser won't do the same thing as XML would if the element
names are in the HTML5 language.
... I believe that the only workaround is to put the XML in a <SCRIPT>
element, that gives you the XML in an escaped node.
MK: Or download the XML separately.
HS: The text node will have the text unescaped.
NW: Oh, OK, yes. If serialized then escaped, but in the node it's not.
NM: The XML need not be for manipulation only in Javascript/CSS, you may
also or instead want to manipulate it in XML (or HTML) tools at the
server, or conceivably elsewhere on a client.
HS: The script element trick works for all languages, so XML is being
treated as a special case.
NM: Yes, and there are arguments pro and con as to whether that makes
sense. HTML and XML have a long history togther, and this task force is
focused on exploring synergies.
JC: Just use XHTML?
NM: Yes, but we always get back to the huge install base that runs best
with text/html
<darobin> "The script element allows authors to include dynamic script and
data blocks in their documents. The element does not represent content for
the user."
NW: I find the uniformity of treatment of all languages by NOSCRIPT to be
appealing.
<Norm> I'm not sure I went so far as to say that I found it appealing, but
...
NM: So, I'm a little troubled by the fact that <SCRIPT> tags have mandated
processing in the case there's a script there. What if the script is media
type applicaiton/xml
JC: Not troubled by that. You'll use something like application/xslt+xml
if you want your XML interpreted as (in this example) an XSLT script.
... Historically, media type is what to do with it, not what it is.
NM: I strongly disasgree with that.
JC: Oh, I mean in HTML
NM: Specifically on the SCRIPT tag
NM: I'd prefer to associate the processing rules with the spec for the
SCRIPT tag
JC: What does the HTML5 spec say?
HS: I agree with Noah that in principle there's an architectural issue; in
practice the set of languages supported in browsers is small and slowly
growing. So far none in XML. If necessary, any such new XML scripting
language could get a more specific type.
Speaking for myself: OK, maybe the HTML5 spec should say what Henri just
said.
JC: XSLT?
HS: They don't support it in <SCRIPT>, and it doesn't make much sense to
do so.
JS: I understand this isn't likely to happen, but not sure why it wouldn't
make sense.
HS: Script processing starts when end tag </script> is parsed, and you
only have a partial DOM. Seems not to make sense to do XSLT then. Hmm, but
a DEFER script could make sense I guess.
JC: Could run multiple successively.
MK: Some of my points have been partly covered. There are a lot of
potential XSLT processing scenarios, many of which can't be captured by
<script type="..xslt type..">
... E.g. when to run, what the input is, whether there's more than one
script, etc., parms, etc.
... Relying on one attribute seems insufficiently extensible. Henri
reinforces that when he says "won't happen in next year, therefore
uninteresting". Seems the wrong way to architect. We should look further
into the future, to when Javascript seems as old fashioned as COBOL. The
world is dynamic.
JC: Propose we add embedded XSLT as another use case.
NW: +1
NM: Too bad we're leaving this behind so quickly. The purpose of our group
is to maximize HTML/XML synergies, and for >certain< purposes XSLT is a
terrific language for HTML scripting
HS: There is some implementation in the runtimes for giving the HTML DOM
as input to XSLT processing (scribe isn't sure he got this right)
... The XSLT program can be put in a script element, and use bootstrapping
Javascript that compiles the XSLT program, and chooses as input tree to
give to that program.
... You can put the output in the DOM.
<Zakim> noah, you wanted to talk about circularity
MK: Yes, we've seen the folks at ETH Zurich do just that, using two
<SCRIPT> elements, one javascript and one XQuery. The former looks for and
runs the latter.
HS: The set of programming languages supported natively by browsers has
always been "1" across multiple browsers, that is Javascript. Internet
Explorer has for years also supported VBScript. There are also good
accessibility(?) APIs that allow languages to be plugged in.
... Gecko allows some extensibility, but for various reasons only for
local content.
... Anyway, the trend is toward focus on Javascript only, and viewing that
as a compiler target for other languages. That said, there is precedent
for having other languages.
... You cannot ever use type="text/vbscript" for data, because there
exists a browser that would attempt to execute it.
BINGO! That's why I don't much like using the <SCRIPT> tag for data.
NW: That is astonishingly unsatisfying. It would make much more sense to
add a new <DATA> element, without the risk that IE would later decide that
type="application/fribble" would launch missles.
HS: The reason it's called SCRIPT and not DATA is that there are only a
handful of elements that don't try to parse their content.
... If we introduce something called <DATA>, it would be incompatible with
the install base of browser.
What about <script type="xxxx" mode="NORUN">?
HS: So, the pattern is formalized in HTML5. An alternative is using
<STYLE>. Another is <XMP>, but that's not hidden by default.
NW: Yeah, I forgot the compatibility problem.
<Zakim> noah, you wanted to ask about NORUN attribute
<hsivonen> existing browsers wouldn't honor NORUN
<jcowan> Announcement: I'm working on a MicroXML parser/DOM called
MicroLark (hommage to Tim's Lark parser from the early days of XML)
NM: I think a new attribute would have fewer problems BUT: I admit that it
would be at best eliminating future problems, and then only rarely. The
advantage would be architectural robustness. It appeals to me
intellectually, but I suspect that even if built it would be used only
sometimes.
NW: We are ADJOURNED.
Summary of Action Items
[End of minutes]
--------------------------------------------------------------------------
Minutes formatted by David Booth's [11]scribe.perl version 1.135 ([12]CVS
log)
$Date: 2011/01/12 21:56:20 $
References
Visible links
1. http://www.w3.org/
2. http://www.w3.org/2010/html-xml/2011/01/11-agenda
3. http://www.w3.org/2011/01/11-html-xml-irc
4. http://www.w3.org/2010/html-xml/2011/01/11-minutes.html#agenda
5. http://www.w3.org/2010/html-xml/2011/01/11-minutes.html#item01
6. http://www.w3.org/2010/html-xml/2011/01/11-minutes.html#item02
7. http://www.w3.org/2010/html-xml/2011/01/11-minutes.html#item03
8. http://www.w3.org/2010/html-xml/2011/01/11-minutes.html#ActionSummary
9. http://lists.w3.org/Archives/Public/public-html-xml/2010Dec/0064.html
10. http://lists.w3.org/Archives/Public/public-html-xml/2010Dec/0064.html
11. http://dev.w3.org/cvsweb/~checkout~/2002/scribe/scribedoc.htm
12. http://dev.w3.org/cvsweb/2002/scribe/
Received on Wednesday, 12 January 2011 21:59:45 UTC