As @zslayton pointed out recently, we generally remember to define handling of null.symbol where symbols are allowed, but the other abnormal values $0 and absent shared symbols are generally (if not universally) overlooked by the specification.
I think the 1.0 spec sidesteps the problem by (a) forcing symbols meaningful to the decoder to be encoded using the system symbol table, and (b) not using symbols in other parts of directives. But 1.1 has much more exposure to such things.
Ideally, we would have a simple, global policy, along the lines of "Any symbols with unknown symbols signal an error when encountered within a directive or module definition." But I'm not immediately sure that's desirable as stated.
Here's some relevant contexts:
[ ] Should arguments to make_string and make_symbol accept values without text?
As @zslayton pointed out recently, we generally remember to define handling of
null.symbolwhere symbols are allowed, but the other abnormal values$0and absent shared symbols are generally (if not universally) overlooked by the specification.I think the 1.0 spec sidesteps the problem by (a) forcing symbols meaningful to the decoder to be encoded using the system symbol table, and (b) not using symbols in other parts of directives. But 1.1 has much more exposure to such things.
Ideally, we would have a simple, global policy, along the lines of "Any symbols with unknown symbols signal an error when encountered within a directive or module definition." But I'm not immediately sure that's desirable as stated.
Here's some relevant contexts:
make_stringandmake_symbolaccept values without text?