Re: New Version Notification for draft-snell-search-method-02.txt

[Austin Wright <[e-mail]>]

I have three suggestions, in this order:

(1) Why not adopt the REPORT method? It seems to be defined with neutral semantics, or at least more neutral than SEARCH:

> A REPORT request is an extensible mechanism for obtaining information about a resource.

This definition sounds like a query, and I like the name: it suggests you’ll provide or receive some sort of report, generated by the server based on the payload.

(2) If we mint a new HTTP method, I favor QUERY.

(3) I’m not opposed to SEARCH, I figure the worst that could possibly happen would be some legacy Java app sees "Allow: SEARCH" and makes a WebDAV query to the resource, triggering an uncaught error due to the unexpected response, causing an infinite loop and overheating the CPU, setting everything on fire.

What I don’t like very much is the name. SEARCH suggests something very specific to me. I think of PATCH, a method where only certain media types make sense.

Austin Wright.

> On Sep 3, 2020, at 10:09, Julian Reschke <[e-mail]> wrote:
> 
> Am 03.09.2020 um 18:04 schrieb Roy T. Fielding:
>> ... > Note that is triggered by the media types, not the data format.
>> 
>> I don't see that as a limitation at all. Those general types don't
>> say anything useful for the recipient. That's why applications of XML
>> are supposed to use a more specific +xml media type instead, and we
>> can require that of future implementations without hindering XML as
>> a payload.
>> ...
> 
> Right.
> 
> Payloads in WebDAV regrettably use application/xml (or text/xml). That's
> hard to change now, but I'm tempted to register several more specific
> types for future use cases (like linking to a PROPFIND payload).
> 
> Best regards, Julian
>