- From: David W. Morris <dwm@shell.portal.com>
- Date: Mon, 10 Jun 1996 18:09:03 -0700 (PDT)
- To: Larry Masinter <masinter@parc.xerox.com>
- Cc: dmk@allegra.att.com, http-wg%cuckoo.hpl.hp.com@hplb.hpl.hp.com
On Mon, 10 Jun 1996, Larry Masinter wrote:
> About the spec:
> >> HTTP/1.1 applications that do not support persistent connections
> >> MUST include the "close" connection option in every message.
>
> Dave Kristol wrote:
>
> > Just a nit, but I think SHOULD is more appropriate here than MUST. We
> > have long agreed that applications must be prepared for the connection
> > to close connections anyway, so there's small advantage to an
> > application's receiving the "close". It's merely cleaner, IMO.
>
> It's not inconsistent with the rest of the spec for us to require
> ("MUST") that senders be strict in what they send (send "close" option
> in every message) even while recipients are liberal in what they
> accept (be prepared for the connection to close anyway). This nit need
> not be picked.
The justification for very late change of persistent connections
from default not supported to default assumed was that there was no change
to the protocol. Removing MUST leaves the client guessing. Requiring
'CLOSE' changes the error recovery philosophy required in the client.
There is no advantage to changing MUST and some disadvantage in terms
of predictable behavior. It should stay.
Dave Morris
Received on Monday, 10 June 1996 18:15:41 UTC