aligning "repeatable requests" language with existing specs and question about repeatability for idempotent methods

[Erik Wilde <erik.wilde@dret.net>]

hello.

https://docs.oasis-open.org/odata/repeatable-requests/v1.0/cs01/repeatable-requests-v1.0-cs01.html#sec_RepeatableRequest

says

"An unsafe request is a non-idempotent request; that is, a request which 
has the potential to change the service each time it is executed."

this is wording that should be changed. "safe" and "idempotent" requests 
are defined by HTTP itself, so this shouldn't be redefined by a spec 
building on HTTP. in HTTP, "an unsafe request is a non-idempotent 
request" is a false statement, because "safe" and "idempotent" refer to 
different things:

https://datatracker.ietf.org/doc/html/rfc7231#section-4.2

using this definition from RFC 7231, it also isn't quite clear after 
reading the "repeatable requests" spec why the proposed header fields 
would have to be used for PUT and DELETE. those methods are defined by 
HTTP to be idempotent, meaning that requests are repeatable by definition.

the "repeatable requests" spec has a more sophisticated model of 
repeatability, but if these header fields indeed should be used with 
idempotent HTTP methods, then the spec would need to carefully and 
clearly talk about how the fundamental repeatability of these methods 
changes by using the additional header fields.

kind regards,

dret.

--
erik wilde | mailto:erik.wilde@dret.net |
           | http://dret.net/netdret    |
           | http://twitter.com/dret    |