In which places is list ordering intended to be significant?

[jdougan]

eg. Is "class":["foo","bar"] supposed mean the same as "class":["bar","foo"] ?
Same thing for entities, links, actions, and fields.

[apsoto]

IMO same

…

On Sat, Jun 24, 2017 at 8:46 PM jdougan ***@***.***> wrote: eg. Is "*class*":["foo","bar"] supposed mean the same as "*class*":["bar","foo"] ? Same thing for *entities*, *links*, *actions*, and *fields*. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#87>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAAtrNpBF2_d5LmLA5OjYkkBbyazqSd0ks5sHdf7gaJpZM4OEgUm> .

[kevinswiber]

From Siren's perspective, there's no priority or significance to the array item position for strings in the class array.

Siren doesn't prescribe how clients should process classes; the spec only defines the semantic meaning. Your implementation gets to make the decision if it's significant.

[jdougan]

I'm not sure how to read that. Either the semantics are such that "class":["foo","bar"] has the same meaning as as "class":["bar","foo"] (collection interpreted as a Set), the semantics say they are not the same (interpreted as an Array), or order must be preserved because it could be important (which is a restricted version of the second). There are no other choices, at least if you care about interoperability.

[kevinswiber]

From Siren's perspective, there's no priority or significance to the array item position for strings in the class array.

1. There's no meaning associated with the contents of the class array.
2. There are no processing instructions associated with how clients should interpret the elements of a class array.
3. There is no composition guidance on how servers should combine multiple elements in the class array.

These are all deliberately outside the scope of the specification.

Prior work in this space includes the class attribute in HTML. Because HTML did not specify detailed semantics around class, CSS Selectors were able to use the attribute effectively.

If we look at SVG, the specification of the class attribute lists the following roles:

• As a style sheet selector (when an author wishes to assign style information to a set of elements).
• For general purpose processing by user agents.

The same is true for items in a Siren class array.

[jdougan]

The W3C does put extra semantics beyond the encoding provided ones for their attributes in general and the class attribute is one of them. Specifically in HTML class, "duplicates are ignored", which appears to be a low-level processing instruction to me. This is the same kind of issue as ordering. Looking at that HTML spec text more closely regarding "set of space-separated tokens" referenced by the class attribute, they very carefully put ordered collection semantics around what is, in the XML model, a raw string. They also carefully defined unordered collection semantics for use elsewhere, as well as taking note that case-sensitivity will be defined elsewhere on a case-by-case basis.
As I (mis?)understand you, if the Siren class attribute was an hypothetically an XML attribute, you would be saying that the all of the interpretation of the raw attribute string would be delegated to the protocol consumers, since that is what the affordance of the encoding gives you.

The SVG spec you reference is unfortunately less clear. However a quick glance shows "set of class names" which may imply duplicates are removed and possibly, if less likely, that order is unimportant. With a quick search, the string "sequence of" is used 21 times in that SVG doc, so it is far from imposssible they are making the distinction, but not conclusive. Looking at the current SVG2 draft to see if it is any clearer, they are deprecating the SVG class attribute in favour of the one on the more general Element, and make specific reference to the WHATWG HTML spec's definition of "set of space-separated tokens".

I think you are misunderstanding the level of semantic clarification I'm trying to ask about. I am thinking about a low-ish level data model, suitable for mapping into and out of other encodings (without just pointlessly writing JSON in the other encoding) and for non-specialized tools for processing and generating Siren. I asked about class because it (incorrectly?) seemed like a simple place to start. I could have asked about class element duplication. I could have asked about ordering in the fields attribute. (For compatibiity with older HTML forms processing, it is probably useful to assume sequence matters there as it far too often does in real systems, under the terms of Postel's Law.) And, as I started this thread with, entities, links and actions are also where I am asking if the sequence is incidental to their representation, or core?