IDL: Make name required

[kenchris]

And look whether other properties need to be so too

[aarongustafson]

The associated PR was merged. We can close this, right @kenchris?

[mgiuca]

But that associated didn't address the issue; this was filed in response to a comment on there.

I think name possibly should be required.

[mgiuca]

Looks like this was already litigated in 2013: #24

I think the issue is that we don't want to drastically error out if there's no name, but we still want to "require" it in what we consider a "valid" manifest. (Similar to how HTML <title> is technically required, but obviously agents still display a document if it's present.)

@marcoscaceres Perhaps you can define this better in the new parser that you're working on. The "correct" thing to do here is to a) define a "valid" manifest, and b) define the user agent behaviour if the manifest is invalid, which may not be fatal; e.g., we may still accept a manifest that has no name but the site may not be installable.

[aarongustafson]

@mgiuca I can get behind that.

[marcoscaceres]

this is in the spec already (at least, option 2 is)... will link it up properly.

[christianliebel]

@marcoscaceres What needs to be done here?

[marcoscaceres]

I still don't think we want to require name, tbh... some UAs may have criteria that invalidates a manifest without a name, but others might not. Safari, for instance, is perfectly happy to let you install a web application without a name.

[marcoscaceres]

s/invalidates a manifest without a name/considers the manifest to not be installable

[aarongustafson]

Without it, users would end up with a less accessible app icon though, no?

[marcoscaceres]

Not necessarily, the name can still be derived from the document's application-name, title, or given a system default like "untitled" (or the user can rename it).

[aarongustafson]

Sounds good. Let's document that.

[marcoscaceres]

It's currently in the authority section:

in cases where metadata is missing, or in error, a user agent MAY fallback to the Document to find suitable replacements for missing manifest members (e.g., using application-name in place of short_name).

But it may be good to actually move the above to the "application name" section.

[mgiuca]

Hmm. #668 which I am trying to solve is working towards having the manifest not depend on the document for processing.

I would rather not spec that the manifest's name falls back to the document. That is even more of a dependency on the document than we have now, which moves us away from being able to say "install this manifest" (without being in a document).

I'm happy to keep a MAY that says the user agent can get stuff from the document. As discussed on #834, that's unavoidable given that user agents invent an entire manifest if installing a page without one, but I want to keep that as a user agent feature, and have the core manifest processing algorithm not rely on a document.

[marcoscaceres]

I'm happy to keep a MAY that says the user agent can get stuff from the document.

Yeah, what's currently there (#622 (comment)) is sufficient IMO. It gives the UA flexibility and doesn't tie it to the document.

[aarongustafson]

Once #927 lands, shall we make this change?