On Sat, Oct 13, 2012 at 2:08 PM, Jan Algermissen <jan.algermissen@nordsc.com > wrote: > > On Oct 13, 2012, at 10:39 PM, Mark Nottingham wrote: > > > > > On 14/10/2012, at 6:11 AM, James M Snell <jasnell@gmail.com> wrote: > > > >>> Right, but what's the difference between: > >>> > >>> Prefer: wait=10 > >>> and > >>> Prefer: return-asynch, wait=10 > >>> > >>> ? "return asynch" really says "give me a 202" which is nonsense; the > client doesn't control the status code, the server does. > >>> > >> > >> That's why it's a Prefer header and not Expect. The server retains > control. "Prefer: wait=10" could just as easily result in the server simply > throwing up it's hands and saying, "sorry, can't do it" > > > > And what does return-async(h) bring to the party? > > I cannot see that either. In fact, I cannot even imagine a server paying > attention to the '10', given the juggling between estimating the task > duration and comparing to the wait time. > > Maybe > > Prefer: wait > > just does the job? > No, wait simply tells the server how long the client expects processing of the request to take and does not communicate any information about the preferred response. I could, for instance, have a theoretical "Prefer: wait=120, respond-giveup" that could be interpreted as "I prefer you give up if you can't process this request in less than 2 minutes". - James > > Jan > > > > > > The server can still throw up its hands with a 4xx or 5xx, and the > client has to deal with that. > > > > > > > > -- > > Mark Nottingham http://www.mnot.net/ > > > > > > > > > >
Received on Tuesday, 16 October 2012 19:34:22 UTC