World Wide Webber - HATEOAS - The Confusing Bit from REST

Comments:

#	great slide deck! Posted by Mike Amundsen, email, url @ 26 May 2009 10:38 UT

#	Your presentations in Falando em Java were fantastic! Congratulations! Posted by Rafael Carneiro, email, url @ 27 May 2009 01:06 UT

#	Congratulations Jim! Your talks were amazing! I hope to see you in Falando em Java Conference next year again! Posted by Rafael Ponte, email, url @ 27 May 2009 01:40 UT

#	Jim! Thanks for your talks... they were just incredible! Posted by Paulo Silveira, url @ 27 May 2009 07:49 UT

#	That was indeed a great slide deck. I have passed this along to several colleagues. Posted by Ryan Riley, email, url @ 28 May 2009 05:23 UT

Jim, I think there's an aspect of HATEOAS that's completely underrepresented in all treatments: How does the client know about the semantics of the links it could follow next.

Commendably, you point out the importance of media types, but then you leave it at that. Is there anything to the media type beyond its unique name? Are client and server just acting on a common (mis-)understanding of what is implied by the media types they are using?

Media types are the 500 pound gorilla in the HATEOAS room that. At best, as in your presentation, its existence and importance is acknowledged. But curiously, not much beyond that is said.

Taking application/vnd.restbucks+xml as an example, how do I specify unambiguously what it means to follow a rel="payment" link?

I assume this is written down somewhere in human, not machine, understandable English and encoded appropriately in the client and server software. If that is the case, then both do in fact know everything about the protocol. There's no protocol left in the hypermedia structure; what this structure still conveys is which subset of the possible transitions is actually available at each point in the conversation. That is important, of course, but only a (small?) part of the complete protocol.

Posted by Michael Schürig, email, url @ 03 June 2009 09:30 UT

Hi Michael,

The media type specifies "look for links in here" which gets us halfway there as you point out.

The resbucks protocol description expresses the link annotations and semantics (it's just a microformat). From that you can build a client which drives the workflow.

On the client side, you use pattern matching (or rules engine, or lots of nasty nested if-else statements) to drive the protocol.

At this risk of being a salesman, we're writing a book about this :-)

Jim

Posted by Jim Webber, url @ 04 June 2009 08:32 UT

Jim,

even if it is "just a microformat" it has to be specified somewhere. And, on top of that, server and client have to agree to use it.

It's likely that I'm just reacting badly to overblown claims for HATEOAS, that I think are often implied. For a slightly lengthier explanation of what I mean, see

http://www.schuerig.de/michael/blog/index.php/2009/06/04/humble-hateoas/

I'm looking forward to the book.

Posted by Michael Schürig, email, url @ 04 June 2009 12:40 UT