Why does `obj.foo = function() { };` not assign the name `foo` to the function?

As of ES2015 (ES6), functions have proper names (including an official name property), and names are assigned when the function is created in a variety of ways in addition to the obvious function declaration and named function expression, such as assigning to variables (function's name is set to the variable's name), assigning to object properties (function's name is set to the property's name), even default values for function parameters (function's name is set to the parameter's name). But assigning to a property on an existing object (e.g., not in an object initializer) doesn't assign that property's name to the function. Why not? Surely there must be a specific reason it was not desirable/possible. What was it?

To be clear: I'm not asking how to work around it. I'm asking what prevents this seemingly-obvious case from being handled when so many others (including default parameter values!) are. There must be a good reason.

Please don't speculate or theorize. TC39 had a reason for not including it. I'm interested in what that reason was. I've been through the TC39 meeting notes but haven't found it yet. The closest I've found so far is Allen Wirfs-Brock replying to Bergi to say there was no consensus for doing it for that form because of "various objections," but sadly he didn't say what those objections were.

Details:

All of the following assign the name foo to the function on a compliant browser:

// Requires a compliant browser

// Assigning to a variable or constant...
// ...whether in the initializer...
{
    let foo = function() { };
    console.log("1:", foo.name); // "foo"
}
{
    const foo = function() { };
    console.log("2:", foo.name); // "foo"
}
// ...or later...
{
    let foo;
    foo = function() { };
    console.log("3:", foo.name); // "foo"
}
// As an initializer for an object property
{
    const obj = {
        foo: function() { }
    };
    console.log("4:", obj.foo.name); // "foo"
}
// Or as a method
{
    const obj = {
        foo() { }
    };
    console.log("5:", obj.foo.name); // "foo"
}
// Even if it's a computed property name
{
    let name = "f";
    const obj = {
        [name + "o" + "o"]() { }
    };
    console.log("6:", obj.foo.name); // "foo"
}
// As a default value for a parameter
(function(foo = function() { }) {
    console.log("7:", foo.name); // "foo"
})();
// ...and a bunch of others

But assigning to a property on an existing object, outside an object initializer, does not:

const obj = {};
obj.foo = function() { };
console.log("Nope:", obj.foo.name);

As far as I can tell, this is covered by this section of the specification, which explicitly only sets the function name if the IsIdentifierRef of the LeftHandSideExpression is true (which apparently it isn't for property references).

So reiterating from above: Why not? Surely there must be a specific reason it was not desirable/possible. What was it?


Allen Wirfs-Brock has replied on the es-discuss list with the objections that prevented the TC39 consensus on the obj.foo = function() { } form:

...for

cache[getUserSecret(user)] = function() {};

it would leak the secret user info as the value of name

and for

obj[someSymbol] = function() {}

it would leak the Symbol value as the value of name

and for

 table[n]=function() {}

name would likely be a numeric string

There are counters to those objections (particularly the last one, which is extremely weak; there are many other ways a function is automatically assigned a numeric string name), but that's not the point; the point is that those were the objections raised.

He also added that the IsPropertyReference operation that would be required (where currently there's just an IsIdentifierRef)...

...is a runtime operation and the new semantics require runtime determination of the name value. This is all extra runtime work that may slow down the creation of function closures that appear within loops.

So all in all, apparently at the time the decision was taken, those objections carried the day (and quite possibly would now as well), which is why this form doesn't automatically name functions whereas so many others do.