Re: interleave and strings

[James Clark <jjc@jclark.com>]

Kohsuke KAWAGUCHI wrote:
> 
> In my understanding, this restriction can be better stated as follows:
> 
> Pattern is illegal if (and only if) its content model automaton has two consecutive
> transitions by characters. I think this is what we want to avoid.

That is not quite it.  I disallow:

<group>
  <string>abc</string>
  <attribute name="foo"/>
  <string>def</string>
</group>

I also disallow

<group>
  <string>abc</string>
  <element name="foo">whatever</element>
  <string>def</string>
</group>

(in order to simplify specification/implementation) but I don't disallow

<group>
  <anyString/>
  <anyString/>
</group>


I also don't think a description in terms of automata is a good for the
spec, because it would require that the spec describe how to build an
automata.  I also want to keep things easy for people who aren't using
automata to implement this.

James