HAL is a lean hypermedia type for your RESTful APIs | Hacker News Reader

HAL is a lean hypermedia type for your RESTful APIs

stateless.co

Reading this leaves me troubled. Why not the XLink vocabulary, for instance? I can understand something like this appearing for JSON if only so that people have an intrametalinguistic transition path for their JSON APIs, but letting so much existing work on interoperable XML representations fall by the wayside could be unfortunate.

Did you actually register the media types?

“Any and all elements are legal in a HAL representation provided they do not conflict with HAL's reserved elements.” If I'm reading this right: please, please, no. Not in XML. Allocate a namespace URI and use it. Let's not start down the path of everyone having to watch out for everyone else's element names again.

The idea of identifying multiple packed related resources (which could otherwise be top-level resources) in the same document is interesting. I don't immediately recall specific prior specifications there, but I'd definitely want to go through the W3C site with a fine-toothed comb to check. E.g., Atom feeds already can embed XML fragments of almost any other type (though most commonly XHTML) as entry content; there may be some evolution of that available.

YAML specifies standard way of including multiple documents in a single stream/file. With XML (assuming XSD-compatible parser), you can always extend the NS and have several documents in a single container (thus single stream/file), but I don't see how that would be preferable to just using MIME/Multipart.

MIME multipart is pretty good in and of itself, and in particular can represent a larger space of content types well. I'm trying to imagine how this interacts with things like URI fragment identifiers, though. MIME has Content-ID, but that's defined to be a msg-id, which doesn't help much. xml:id might be more effective in a sense, but then you run into the internal collision problem again.

Implementationally, hooking up an extra MIME layer is a lot more cumbersome than it should be. This is theoretically a transient characteristic, but I have the creeping feeling that that will destroy any attempts to use MIME for this purpose regardless.

Well, yes, if the content is inter-connected this is more tricky.

Xml:id assumes unique identifiers within scope, and with multiple documents that would require a global scope, something that is not standardized and problematic (it'd require either using UUIDs as IDs or keeping track of uniqueness in some other way).

I think a better solution would be to use XLink. Assuming each document has its own canonical link, this is then straightforward:

1. Make sure each XML document within MIME/Multipart container includes <link rel="canonical" href="uri:…"/>

2. Each document in a MIME/Multipart container should refer to other documents using xlink, i.e. unique URL with a part identifying specific fragment using XPath (which can then use xml:id in a local scope).

Now, depending on the client following scenarios are possible:

1. Naïve: Client will perform redundant requests, but will cache documents in containers, resulting in several HTTP(304) responses.

2. Smart: Client will notice that referred documents are contained in a multipart MIME message and will use local versions.

As a bonus, this approach is not dependant on MIME in any way — it will work just as well with, for example, a ZIP archive.

Might be worth exploring if re-using stuff like XLink makes sense and is within keeping of the objective to keep things simple. Probably better to discuss that on the google group than here, feel free to join and raise it there.

At the moment, it's registered with IANA as application/vnd.hal+xml, although I don't think it makes a huge difference either way at this point time.

Element collision is not a big deal as clients should be determining element semantics on a per-resource basis and - importantly - in the context of the link relation leading to it (from either a resource or a link).

You will find similarity with other media types like atom and rdf. Basically, it's designed to be more generic than atom and less convoluted than RDF.

No mention of IANA's link relations[1], OpenSearch[2] and Atom[3] which seems to cover almost anything and beyond what ‘HAL’ specifies.

For JSON, JSONSchema specifies standard way of including hyperlinks in a document, but of course no one, that uses JSON in the first place, even cares.

[1] http://www.iana.org/assignments/link-relations/link-relation...

[2] http://www.opensearch.org/Home

[3] RFC 4287, RFC 5023, RFC 4685, RFC 4946, RFC 5005

Having pretty much sworn off XML, at least for my own projects, I've been pondering what JSON should look like that if you are going all the way to HATEOAS and you have links in your data returned from your REST interface.

You will reinvent XML with uglier syntax, all while hurting yourself and Open Web. I've managed to somehow get into that argument before, so maybe there's something worth reading there: [1], [2].

[1] http://news.ycombinator.com/item?id=2623824

[2] http://news.ycombinator.com/item?id=2588606

Don't get me wrong - I'm not an XML hater by any means it's just that JSON appeals to be as a syntax - possibly as I did Lisp development for a number of years and the relative similarity of JSON to s-expressions appeals to me.

Well, I only care for public Web APIs. Though if you're not writing your documents by hand, I don't see how the syntax matters (it's just library then, anyway). And if you are writing your documents by hand, there are better solutions than JSON — for example YAML (which is excellent!).

Yep, it definitely should be making reference to the IANA registry (will likely do this by referencing Web Linking RFC5988).

The point here was not to necessarily invent new capabilities, but to take a unique position in its design and bring together the most consistently useful assets of existing types whilst filtering out the noise.

It's much healthier for the web than just returning plain JSON/XML or a custom media type.

It's preferable to return correctly annotated (with elements from aforementioned namespaces) XML, rather than ‘plain XML.’ My point was, that this vocabulary already exists and there's no reason to invent new (incompatible) one. If you think you can provide some improvement to any of that, you should definitely contribute to that standards.

It's great that you care about what you output at all. Most developers don't, some even try to embrace their ignorance. I think that there's certainly a lack of user-friendly resources, that introduce developers to the best practices regarding data formats and public APIs. For example, a website, that would catalogue and provide examples of usage of certain conventions, would be, undeniably, a godsend.

I hope we can all make the Web a better place, and just adopting already defined standards would be a tremendous achievement!

HAL already incorporates standards for the URI template spec and the CURIE syntax. It's also going to be rejigged to incorporate terminology direct from the Web Linking RFC.

The above being the case, it doesn't seem as if HAL is re-inventing anything significant from a standards perspective, and avoids inheriting baggage associated with pulling in elements from other namespace. Perhaps not a completely 'pure' approach, but the primary objective is simplicty - and I think doing things this way is a good balance.

If this is something you feel strongly about, please join the google group and raise it there so we can discuss merits and trade-offs in more depth.

Ah man. Where was this when I was making my latest REST service!?

Ah well. I'll have to remember it for the next one.

yeah.. sorry about that ;)

Sketched it out about a year ago, the design has changed a tiny bit since then but it's basically the same http://bit.ly/acgCyY

Here's a nicely done comparable: http://json-ld.org/ Compatible with RDF while keeping it simple.

Not a single mention of RDF throughout. Am I missing something, or is this wheel reinvention?

The intention with hal was to keep things as simple as possible. RDF (and its related media types) don't do a very good job of this, in my opinion. I think the real-world usage is a good indicator of whether RDF is appropriate for most people's use cases. There's probably some node.js analogy to be made here, too :)

Reply on news.ycombinator.com