Clarify+make fitness distance algorithm agnostic to dictionary member presence. #526
Conversation
getusermedia.html
Outdated
| 0.</li> | ||
| <li>If no ideal value is specified (<var>constraintValue</var> | ||
| either contains no member named 'ideal', or is itself a bare | ||
| value and bare values are to be treated as 'ideal'), the |
There was a problem hiding this comment.
grammar error here - the second part is in the positive tense and should have been in the negative tense - may be better said as "either contains no member named 'ideal' or, if bare values are to be treated as 'ideal', isn't a bare value".
|
@alvestrand Had to add more. Can you take another look? This is to allow whatwg/webidl#76 removing the invariant that member dictionary members are always present. This means we have to test for |
Correct.
What? Why? |
|
@bzbarsky You're right, you explained to me that null is a valid value for an IDL dictionary type and gets converted to empty dictionary. Will fix. |
| @@ -4114,15 +4114,19 @@ <h2>Methods</h2> | |||
| value being compared against.</p> | |||
| <p>We define the <dfn>fitness distance</dfn> between a | |||
| <a>settings dictionary</a> and a constraint set <var>CS</var> as | |||
There was a problem hiding this comment.
So a constraint set is a webidl dictionary, right? That's not really specified anywhere.
Also, https://w3c.github.io/mediacapture-main/getusermedia.html#dfn-selectsettings should make it clear that it has a constraint set argument, not an ambient constraint set.
| <var>CS</var>, of the following values:</p> | ||
| <ol> | ||
| <li> | ||
| <p>If the constraint is not supported by the browser, the | ||
| fitness distance is 0.</p> | ||
| <p>If <var>constraintName</var> is not supported by the |
There was a problem hiding this comment.
So this is a little odd, actually.
The input to this thing is a WebIDL dictionary, right? If a browser doesn't support some random spec that defines some constraint, it wouldn't know to include that name in the dictionary to start with, so it would not trigger this clause.
Is this clause meant to just apply to constraints that the browser knows about in the sense of having them in its IDL but doesn't actually implement matching for? Or is it meant to apply to constraints that the browser really doesn't know about? In the latter case, the input here obviously can't be a dictionary, because those presuppose a fixed set of known members and ignore everything else.
What do UAs actually implement here?
| 0.</li> | ||
| <li>If no ideal value is specified (<var>constraintValue</var> | ||
| either contains no member named 'ideal', or, if bare values | ||
| are to be treated as 'ideal', isn't a bare value), the |
There was a problem hiding this comment.
This looks wrong to me. It's saying that if bare values are to be treated as ideal and constraintValue is not a bare value then the ideal value is not specified. That's clearly not the intent, but that's what it says.
I would strongly urge clarity of processing model over brevity here. If we have an "ideal value is specified" boolean with a complicated way to compute it, we should compute it in some substeps instead of trying to jam the computation into a parenthetical. Code that does that is not considered a good idea; neither are specs that do it.
| <li>If no ideal value is specified (<var>constraintValue</var> | ||
| either contains no member named 'ideal', or, if bare values | ||
| are to be treated as 'ideal', isn't a bare value), the | ||
| fitness distance is 0.</li> | ||
| <li>For all positive numeric non-required constraints (such as |
There was a problem hiding this comment.
What does "numeric" mean here? Is that a property a constraint can have as part of its definition (in which case please link to the concept's definition) or something else?
| @@ -4185,7 +4189,7 @@ <h2>Methods</h2> | |||
| <li>Each <a data-lt="constraints">constraint</a> specifies one | |||
| or more values (or a range of values) for its property. A | |||
| property MAY appear more than once in the list of 'advanced' | |||
| ConstraintSets. If an empty object or list has been given as | |||
| ConstraintSets. If an empty list has been given as | |||
There was a problem hiding this comment.
It's really not clear to me what cases this is supposed to be handling. Is this dealing with constraints that are sequence-valued?
Fix for #525. @bzbarsky PTAL.