rewording of title and language

[mattgarrish]

Here's an attempt at formalizing the wording from issue #48.

I'm not completely comfortable with it because we're flirting around with serialization, but hopefully it passes muster as a cleanup attempt.

---

Preview | Diff

[iherman]

I have a growing concern that is, I admit, a bit theoretical and/or purist. In general, the various type of metadata, like the title or the language tag, in an HTML file is meant for that HTML content only. What we are doing here is, in some sense, misuse that mechanism by declaring that, say, the language tag is for the whole WP, ie, for all the resources. In other words, the effect of that language tag is valid for other HTML contents, for example (unless explicitly re-defined in that resource). It is fine as a fallback, but I would somehow think that a stronger warning around this should be put there. Maybe, at least for the language tag, we should not say that the language tag MAY be omitted but, rather, that if it is omitted, the UA may choose to use that. Ie, put it in the bullet items below.

I am not sure the same holds with 'title'. If we consider the first entry in the reading order as, potentially, the cover page, then it is sort of fine.

[mattgarrish]

Well, I don't really like harvesting information from other sources, as I've said before, and you're hitting on why. That's why I prefer there be explicit authoring intent for doing so, and preferably this kind of harvesting should only occur when the manifest is embedded in another document where we know this information applies.

Pulling information from random sub-resources will net "something", but the value of that something is in question in my mind. I'm okay with it for failure handling, but we're still playing loose with when an author has done something intentionally.

I did put a note trying to explain that this language has no effect on processing. Originally I had it as a caution, and will update back to that. I'll also try to beef up the explanation that this is informative metadata as far as processing goes.

[HadrienGardeur]

I share your concern @iherman and @mattgarrish about publication vs resource level metadata.

We can rely on resource level metadata for the publication as a fallback in failure mode, but we'll need to make it very clear that this is not how publications should be authored.

[mattgarrish]

I think what I'll do is change title the same way as language.

I think we're drifting into projecting serialization, and while that has to happen, I don't think we can know how implicit harvesting will work until we have more details.

We can always return to this problem and adjust later.

[HadrienGardeur]

If we have a canonical identifier, it probably makes more sense to use the canonical identifier instead of the address in link@rel="identifier".

The identifier value is just a proposal at this point, not sure that we should talk about it at all.

[mattgarrish]

Could you open as a new issue? I'm working on consolidating the three PRs and reverting the defs so will be closing these off shortly.

[mattgarrish]

Closing in order to re-open consolidated PR. See #51.