[Catalyst] Re: In search of RESTful CRUD holy grail

[Aristotle Pagaltzis]

* Zbigniew Lukasiak <zzbbyy at gmail.com> [2009-03-04 21:45]:
> OK - does that mean that you agree that someone can build a
> RESTful CRUD with that URI structure?

Yes, because one can do with almost *any* URI structure. :-)


* Peter Karman <peter at peknet.com> [2009-03-04 23:25]:
> When I hear "RESTful URL" I interpret that as "URI-as-a-noun"
> (or more negatively-defined, "URI-with-no-verbs").
>
> Perhaps we should use something like "REST-friendly URL"?
>
> [0] e.g.,
> http://www.mail-archive.com/catalyst@lists.rawmode.org/msg01712.html

Yeah. And URIs-as-nouns is a good goal that should be encouraged,
absolutely. URIs should identify resources, not operations, that
is true.

However, putting too much emphasis on the structure of URIs is
at the same time problematic, because while it drives people away
from a terrible style, it also pushes them in the direction of
another bad style (however a less problematic one, to be sure).
They forget to put links into their output documents, instead
putting application-specific IDs in there, and then they publish
API docs that consist of tables of URI construction rules. Then
clients have to be hardwired for the interaction between those
IDs and the URI construction rules.

Which is exactly the opposite of REST.

In REST, URIs are opaque as far as the client code is concerned.
Clients only ever know what URIs to follow because they get them
as links or build them from forms or URI templates they find in
the content of server responses – *at runtime*. Not hardwired in.
The documentation of an actually RESTful API explains the format
of the server response and where to find links in it, as well as
what methods can be used against what type of link, and for what
purpose. The server URI structure is totally irrelevant. If it
changes tomorrow, and then again the day after, and again next
week, you don’t need to care; the response format is the same
(or actually, gets extended in compatible ways) so you can still
find the links in the right places and the client magically keeps
working while the server completely changes its implementation
and turns its URI structure upside down. *That* is the point of
REST: completely decoupling servers and clients from each other.

Good URI design is useful as a usability issue for humans OT1H
(bookmarkability, hackability etc are usability concerns), and
as a server implementation concern OTOH. But it’s entirely
orthogonal to the principles of REST.

Regards,
-- 
Aristotle Pagaltzis // <http://plasmasturm.org/>