[tei-council] content model of <surface>

[Sebastian Rahtz]

In order to give <surface> a cats chance of hell of getting a deterministic content model,
I have changed it to be:
 
    <zeroOrMore>
	<choice>
	  <ref name="model.global"/>
	  <ref name="model.labelLike"/>
	  <ref name="model.graphicLike"/>
	</choice>
      </zeroOrMore>
      <zeroOrMore>
	<group>
	  <choice>
	    <ref name="zone"/>
	    <ref name="line"/>
	    <ref name="surface"/>
	    <ref name="surfaceGrp"/>
	  </choice>
	</group>
	<zeroOrMore>
	  <ref name="model.global"/>
	</zeroOrMore>
      </zeroOrMore>

where the change is that it no longer allows all of model.glossLike, only
model.labelLike at the start. This means we allow <label> where it was not allowed
before, and lose <altIdent>, <equiv> and <gloss>. 

So surface now allows a pretty promiscuous set of zone/line/surface/surfaceGrp + model.global (which includes
the certainly elements), but if you want <desc> or <graphic>, they must precede any of the zones or lines or
sub-surfaces.

I claim this is what the geneticists want.


PS why did all this come up, you ask? because I noticed that "testall" (== tei_all) had
been omitted from the normal test regime, and when I put it back  this non-determinism
revealed itself.


Sebastian