Re: [w3c/manifest] Update manifest spec to consider manifest updates for icon urls to be cache-control:immutable (PR #1199)

@dmurph commented on this pull request.



> @@ -1792,6 +1792,13 @@ <h3>
             `name_localized`.
             </li>
           </ol>
+          <p>

>> User agents MAY consider an icon unchanged if the user agent determines the visual difference is insignificant.

> I thought this was an implementation detail of Chromium and not to be speced. This was also not proposed in the explainer, does it make sense to spec it? I applied your changes nonetheless, kept this as an open question.

The spec tries to be clear to both devs and user agents about what behavior should be expected and allowed. The first sentence is a contract for the user agent - if a url changes, the developer expects the user agent will be checking new urls for icons. However, this is not the full story - can the developer expect that this is the ONLY time a user agent might check an icon? No. So the second sentence communicates to devs that the icons can also be fetched in other circumstances, and gives flexibility to user agents to do so (like we do sometimes).

cache-control immutalble also specifies this flexibility, with some specific exmaples - if you want prior art for this happening

>> Spec is binding, and our implementation also considers icons if any of them change.

> Same question as above, I thought this was an implementation detail and not something we have proposed in the explainer, which [only suggests icon url changes](https://github.com/WICG/manifest-incubations/blob/gh-pages/predictable-app-updating.md#proposal). Should we still spec it according to implementation?

Unfortunately it looks like your old commit is gone (did you amend?) so I cannot see your old language. From my memory, the language was written in a way that would make our implementation non-spec-compliant. It was incorrect.

>> IDK if we need to mention cache-control immutable
> Imo, it might be nicer to keep the explainer, PRD and the spec say the same thing. It makes it easier to follow the chain of research and understand why it was kept that way. Wdyt?

That is a good way of 'understanding' the concept. We could add a 'non-normative-note' section for that. But to formally describe the behavior of this in terms of that spec, we have to hook this into concepts and algorithms in there, which seems difficult. You might be able to ask another spec person like Marcos about exactly how to do this, but I'm not confident enough in my knowledge to do this in a way that is correct and also is understandable / thorough.

I can see having a non-normative note here helpful about cache-control immutable.

-- 
Reply to this email directly or view it on GitHub:
https://github.com/w3c/manifest/pull/1199#discussion_r2603972676
You are receiving this because you are subscribed to this thread.

Message ID: <w3c/manifest/pull/1199/review/3559211971@github.com>