What if encoding is not an EncName?

The new PR adds the following text: "The value of this attribute is an 
EncName as defined in XML specification, section 4.3.3, rule [81]"

OK. That seems eminently reasonable. Looking in the XML spec I find:

EncName	   ::=   	[A-Za-z] ([A-Za-z0-9._] | '-')*

No surprises. Ann encoding name contains ASCII alphanumeric characters, 
period, underscore, and hyphen. Nothing unexpected here.

However, what should I do if the encoding attribute does not satisfy 
this production? e.g. it's encoding="ISO 8859 1"?

Is this a resource error? A fatal error? Should the processor simply 
ignore the encoding attribute? or treat it as if it's an unrecognized 
encoding?

I don't know that this requires going back to CR, but it should be 
clarified. Also there should probably be a test case for this.

-- 
Elliotte Rusty Harold
elharo@metalab.unc.edu