Design goal regarding HTML5

On Tue, Sep 4, 2012 at 10:30 PM, John Cowan <cowan@mercury.ccil.org> wrote:

> One of the nice things about MicroXML documents, indeed
> James's original motivation for them, is that they are a good format
> for human readable documents.  You can keep them around for processing
> with XML or MicroXML toolchains, but if they use the right elements
> and attributes, you can also view them casually in any web browser.
> The only price of that is a fifteen-character signal (sixteen if you
> have a trailing newline) that forces the web browser to DTRT rather than
> enabling some unpredictable quirks mode.
>

I think we badly need to get some precision about our design goals are with
respect to HTML5.

My original goal was that you could author valid HTML5 documents by

- following the syntactic rules of MicroXML, and

- following the constraints of the HTML vocabulary, where these constraints
were expressed in terms of the MicroXML data model.

MicroXML would thereby provide a way to avoid the syntactic complexity of
the HTML syntax of HTML5.

However, there are a number of things that stop this working:

a) information about the DOCTYPE declaration is not in the MicroXML data
model

b) HTML5 doesn't treat <X/> and <X></X> as equivalent

c) MicroXML doesn't allow prefixed attribute names, but some attributes in
HTML (notably xlink:href in SVG) require prefixes

In addition, I'm not sure that many people found my original goal
particularly useful.

Point (c) is also a problem for out current design goal:

8. MicroXML shall be able to straightforwardly represent HTML

So at this point I'm not at all clear what we should be trying to achieve
vis-a-vis HTML5.

If the goal is to be able view MicroXML documents that happen to use the
right element/attribute names "casually in a browser", then does it matter
that the browser will use quirks mode (if we don't allow DOCTYPE)?

For robust, standards-compliant delivery of HTML5, my current inclination
is to use an MicroXML->HTML5 serializer that does the right thing with
empty elements, and adds a DOCTYPE  declaration. We should at least have a
standard way of representing an HTML5 DOM (at least any HTML5 DOM
representable in the HTML syntax) in MicroXML, which means we need some
convention for dealing with xlink:href.

So my current position is:

- no bare DOCTYPE
- no to the additional HTML5 restriction on XML comments

James