Re: I-D ACTION:draft-ietf-webdav-quota-01.txt

> To help us in our categorization of errors, could you describe
> how does a web spider responds differently to a 4xx or a 5xx?

They have different correction procedures.  If you get a 5xx error,
header fields like retry-after become effective and further errors
are directed to the server administrator.  A 4xx error, OTOH, is
directed to the page owner of the referencing page or mechanism
that generated the request.  A 5xx is never seen as a failure of
the author of a reference.

> Also, although it is good to know that one way is compliant and the
> other is not (:-), it would be helpful to have some criteria to
> to actually make that determination.  In particular, how do you
> distinguish a client error category (making a bad request) from a
> server error category (not satisfying a valid client request).

If the same identical request might succeed were it not for a
limitation on the server handling the request, then it is a 5xx.
Basically, there is nothing wrong with the request.

> There are some obvious
> cases (a syntactically ill-formed request), but many of the others
> could fall in either category (the quota errors, in particular).

Not really.  If the request would have succeeded but did not due
to an addition of bytes exceeding quota, then it is a 5xx.
If the request alone is sending a body greater than the quota max
(disregarding other content on the server), then the correct
response is not quota-exceeded but rather 413 (Request Entity Too
Large).

I haven't encountered a case that wasn't obvious.  I have encountered
many cases where extensions have tried to say too many different
things with one status code when two would be clearer.

....Roy

Received on Tuesday, 25 March 2003 19:07:18 UTC