Personally in your example I very much like notation "update_item.id", because there is a clean signal so "id" is the function's argument. When you use "$id", then it is not clean if "id" is a local variable or function's argument. So your proposal decreases safety :-/. Plus this syntax reduces collision only on one side, you should use aliases for sql identifiers and again it is not balanced - In MS SQL I can write predicate id = @id. But it is not possible in your proposal (and it is not possible from compatibility reasons ever).
I understand that there can be problems with functions with very long names.
Right. The problem is most pronounced when the function name is long. And in particular when there are a lot of optional arguments. In this case the caller will be using named arguments so I want to avoid prefixing the parameter names. And while alias is an option, that can also get quite verbose when there are a large number of parameters.
So I think introducing new syntax is not necessary. The open question is a possibility to do aliasing more comfortably. ADA language has a possibility to rename function or procedure. But it is much more stronger, than can be implemented in plpgsql. Probably the most easy implementation can be a possibility to specify a new argument's label with already supported #option syntax.
CREATE OR REPLACE FUNCTION very_long_name(par1 int)
