[Mailman-Developers] REST API PUT versus PATCH - is there in any point in PUT and PATCH?

[Barry Warsaw]

On Feb 19, 2015, at 08:10 AM, Andrew Stuart wrote:

>Is there any actual benefit in those PUT methods?  I can’t see what they
>offer over the PATCH methods.

PUT is defined in HTTP/1.1.  PATCH is (only) defined in RFC 5789, so it may
not be universally supported.  E.g. back in the restish days, I had to hack
WebOb to support it, and it was an ugly hack, though Falcon supports it out of
the box.  I haven't dug into HTTP/2 enough yet to know whether PATCH has
finally been adopted into the official standard.

Much of the philosophy of the REST API was influenced by RESTful Web Services
(Richardson & Ruby), which IIRC recommended implementing both methods[*].  I
recently received the follow up to that excellent book, RESTful Web APIs, but
haven't really had time to read it yet to see how the intervening 6 years have
changed their recommendations.  A *very* quick skim of the first index hit
provides some additional rationale as to why both are useful.

* PUT is idempotent.  If you send the same PUT request 10 times, the result is
  the same as if you'd only sent it once.

* PATCH is not idempotent, but it can save on bandwidth and complexity.  It's
  like a diff, in that it only contains the parts of the resource that have
  changed.

Cheers,
-Barry

[*] Leonard Richardson was a colleague of mine on the Launchpad project back
in the day.  We had many conversations about Launchpad's REST API and his
insights, expertise, and experience deeply shaped our current design.