Hi Mike:

Let me see if I got this right: in a hypermedia design I need to come up with some rules, which can be roughly grouped into three distinct sets:

- rules about how to structure or format my messages
- rules about how to exchange my messages over a network
- rules about what my messages mean in the context of my application or problem domain

As you can see, I found it easier to think about sets of rules than "levels of semantics", which was a hard concept for me to grasp.

Cheers,

Ferenc

Mike:

I quite obviously cannot prove that the implementation exhibits the appropriate architectural style without discussing architectural analysis and design. What I meant is simply that a design may be appropriate for its intended purpose without fully conforming to a named and described architecture style like REST.

I mean by application design a concrete implementation of an architecture, and by architecture an concrete implementation of an architectural style. Or if you wish, that my interest in "appropriateness" lies two levels of abstractions below Fielding's.

I cannot reasonably hope that by simply picking the correct architectural style my design is guaranteed to.turn out correct. I have the strong suspicion that this is the kind of reassurance people who are asking "Is my design RESTful?" are looking for.

I could have stopped reading at the second decision point, but read until the end nevertheless. I would move this second decision point to the first place, because the practical aspect of "appropriateness" seems more important to me than the more abstract and theoretical discussion about architectural styles. The way it is I was tempted to stop reading at the first decision point, because I'm not really interested in discussions about architectural styles as such. Now you may argue that the "appropriateness of design" question cannot be answered without discussing the latter, but I'm not entirely convinced that this is the case.

"I am arguing _against_ using direct serialization of private objects and sending them over the wire". Yes, that was clear, and also that existing tools don't "level up", which I take to mean that they encourage exposing implementation details, for instance database primary keys as JSON object ids. What private object means (object in an implementation language, a database record?) and what is a standard representation (a standard, perhaps domain-specific, media type?) is were I have to fill in the blanks and wonder if I do it correctly.

"Representation abstraction" in the sixth paragraph links to "HTTP 1.1 content negotiation". If it is content negotiation, why don't you say so? If it is not, why do you link to it? I wouldn't be nitpicking if your point were easy to understand. It looks like you are arguing for the use of standard instead of custom media types (serialization) and that you see the biggest impediment in the lack of proper tool support. I'm not sure I got it right, however. Did I?