Re: [core] Robert Wilton's No Objection on draft-ietf-core-problem-details-05: (with COMMENT)

[Carsten Bormann <[e-mail]>]

Hi Rob,


> No real objection except that it feels like this is yet another part of the
> "HTTP stack" that is ported into COAP in a somewhat bespoke way.  

Without the background that led to the current push to finish this document quickly, it might look this way.  The reason we want to finish this now is a specific request from 3GPP people who want to use this in their CoAP APIs in Release 17 (i.e., now).

> I'm not
> convinced that the on the wire byte savings will end up being that great,
> considering that there still seems to be a reasonable amount of text returned
> in the error response.  

Actually, Section 3.1.1 shows an example where there may not be much need for text, but a simple problem details data item of usually 3 bytes may be all that is needed (add 3 more bytes for including the Content-Format in the response).

> Further, although the responses end up being machine
> readable, I'm not sure how easily peer software can really automatically handle
> those errors other than the error indicating that it is a transient response
> and hence potentially the operation could be retried.

The response code will tell the application whether this is a transient server-side problem (5.xx, for which the client indeed cannot do much except for waiting and retrying) or a problem that can be influenced by the client (4.xx).
The example in Section 3.1.1 also happens to be an example for an obvious reaction: retry the operation without the offending option.

Grüße, Carsten