LOL Friday afternoon, I’m tired… time for a rant!
This is a follow-up to my post on the seeds of REST’s destruction.
My understanding of REST improves on a daily basis. Had I written that post today, I would cite an even more fundamental problem that makes moot all the discussion above. That problem is, a definition of a REST service really needs to start by identifying the hypermedia type(s) involved. But there isn’t a hypermedia type available that’s appropriate for general purpose service-to-service communications. Atom is good in the blog world, but there isn’t something like HXML or HJSON that corresponds to HTML.
I’ve cited earlier Roy Fielding’s 2008 blog post, where he blasted a few specs
for NOT being REST, and actually provided some clear principles about
what an API needed to be REST. One of the key points was that without
hypermedia, it’s not REST.
Which basically means that in the ensuing 2 or 3 years since that post, practically no
one has actually started doing REST, and the ones who are, aren’t
getting nearly as much attention as the pretenders.
If his post came out last month, well, sure. Standards take a long
time to propagate. But aside from the investigation you’re doing with
hypermedia, and the Atom guys, I just don’t know that the industry has
done anything to make it a reality. Instead, they cite the three-level
model of REST maturity, ignoring that only level 3 is really actually
Next we’ll have the three-level model of pregnancy, where the first
level is, you talk about being pregnant….
Now, there’s a lot of work going on around HTTP-RPC. What I used to call HTTP/XML or HTTP/JSON before REST became a hot commodity. But if you’re just serializing objects and transferring them using HTTP, you’re not doing REST, even if you pay really close attention to how you build your URIs and pay really close attention to nouns and verbs.
So I don’t think the question is really, “when will REST come to an end”, but “when will it start“?