Reflecting as non-empty URL should special-case empty string

[zcorpan]

https://html.spec.whatwg.org/#reflecting-content-attributes-in-idl-attributes

If a reflecting IDL attribute is a DOMString attribute whose content attribute is defined to contain a URL, then on getting, the IDL attribute must parse the value of the content attribute relative to the element's node document and return the resulting URL string if that was successful, or the empty string otherwise; and on setting, must set the content attribute to the specified literal value. If the content attribute is absent, the IDL attribute must return the default value, if the content attribute has one, or else the empty string.

From web-platform-tests/wpt#2663 (comment)

A number of attributes are defined to take a non-empty URL, and they do nothing when the value is empty; e.g. <img src>, <track src> (but not e.g. <a href>). But when getting the IDL attribute it still resolves the empty string into a URL, which doesn't make any sense. I think we should fix reflecting to have different behavior between <img src="">.src and <a href="">.href such that the former returns the empty string.

[annevk]

This seems reasonable to me, provided at least two implementers are willing to go at it.

[zcorpan]

cc @tkent-google (you filed web-platform-tests/wpt#2663) @sicking (you had opinions in https://lists.w3.org/Archives/Public/public-whatwg-archive/2010Mar/0067.html) -- are you OK with implementing this? I can fix the spec to address this issue and write new web-platform-tests to check this across all relevant elements, etc.

[sicking]

Agreed, we should make the attribute reflect the actual used value. So for elements where "" is treated as "no value was provided" we should make the reflected attribute indicate that.

I'd say that more often than not, treating "" as "no value was provided" is the more expected behavior by developers. It'd be nice if we were somewhat consistent about that.

[foolip]

It'd be nice if as many URLs possible have this behaviour, and that the current behavior is the exception where required for legacy. Otherwise the current behaviour will probably remain the default in bindings generators, and be very likely to spread to many new contexts by accident.

[tkent-google]

I'm ok with the change.