Waldek,Thanks for the outstanding explanation - particularly the significance of the distinction between LL and LR parsing. I have never seen such a clear explanation before. After about 3 reads I understood the whole thing.
I arbitrarily distinguish lexical tokens as a separate entity because tokens like := need character lookahead at what I term the lexical level. At the "tokens" level, tokens lookahead is needed.
The new-ordinal-type production can be resolved if the LL parser has access to an identifiers symbol table. The first identifier of the enumeration will always be undefined. Any subrange identifier must have been defined previously.
I have not noticed any direct left recursion in the 10206 grammar. I will hope that there are indirect LL recursion traps awaiting.
I will proceed with the assumptions that identifier type information will be similarly useful in other cases and that determining the "type" of an identifier will never be ambiguous in an LL sense.
Thank you for taking the time to write such a comprehensive reply.
Regards,
Paul