- From: Shel Kaphan <sjk@amazon.com>
- Date: Wed, 30 Aug 1995 20:46:54 -0700
- To: http-wg%cuckoo.hpl.hp.com@hplb.hpl.hp.com
] > And the wording in section 6.2.2 should change to say that the response ] > "should" include a Location header field, instead of "may", when the ] > entity in the response corresponds to a resource. ] > ] I don't know -- I think you should look at the uses of "should" in ] 8.19 a little more closely. I'm not sure what your objection is. My objection is that there is no reason to require Location headers unless the server has a reason to include them. The current spec does not require them except in a few specific instances (like 302 responses, which I think is the only required place), and I don't see any reason it should be changed to make them required. But, if a server doesn't include a Location header field in a response to POST, then you're shopping basket application won't work, just as you seemed to be complaining about. That's right, but not all applications are the same as my example. The point is that a server wanting to implement an example like mine can do so, and one that is doing any of an infinite variety of other things doesn't have to behave the same way. So it seems to me you should be in favor of a stronger wording. No, I'm in favor of using things where they make sense and not otherwise. As to the meaning of should, lots of RFCs use the following definition for "should": This word or the adjective "RECOMMENDED" means that there may exist valid reasons in particular circumstances to ignore this item, but the full implications should be understood and the case carefully weighed before choosing to do so. Thus, "should" shouldn't be used if one of the exceptions is (e.g.) "when using POST" -- i.e., a very common thing. Paul I think, as you suggested, you may be reading some other document than the one I'm reading, as all the uses of "should" in the various descriptions of the Location header I have only refer to what the URI in the header "should" be like, not when the header itself "should" be present (except the discussion on 302 responses, which is uncontroversial) --Shel
Received on Wednesday, 30 August 1995 20:51:34 UTC