I wonder how many people care about HTTP methods anymore. Tools like
GraphQL seem to just shrug and POST all the things. I
kinda get it. Having a POST body makes it easier to send
complex queries to the server. I’ve been struggling a little bit at
work to devise URL parameters for specifying a bounding box on a map,
or ranges of values that could either be vectors or scalars.
Part of me thinks that maybe the HTTP methods aren’t really
that important. It’s mainly convention, right? As long as
your client and server agree on the meaning of the HTTP methods, it
should be fine. Maybe it doesn’t matter that your
PUT method is not idempotent, or that you use a
POST just to retrieve data. Or if you just want to
POST every request just so you can use that sweet, sweet
post body and encode the actual method as a property of the body;
well, where’s the harm?
This doesn’t feel like it has the same kinds of consequences as
ignoring something like the HTML standard. When you decide that you
can just make everything a <div>, you create tons
of problems for people relying on assistive technology. I can’t think
of any adverse consequences of misusing HTTP methods.
The HTTP methods are useful, I suppose, because browsers implement
them. If I make a <form method=GET>, then I know
that the browser will update the URL the form values. Handy if I want
the pages to be bookmarkable and sharable.
Being able to lean on The Platform™ like this seems like the main benefit of continuing to use HTTP methods. Of course, I could use any method and my server could just redirect to a page with the form values in the URL. It might translate some of the query parameters into path components in the URL, if that’s preferable.
Personally, I always feel icky using a POST request to
fetch data, but maybe that’s an antiquated attitude these days.