Peter Zehler
XEROX
Networked Products Business Unit
Email: Peter.Zehler@usa.xerox.com
Voice: (716) 265-8755
FAX: (716) 265-8792=20
US Mail: Peter Zehler
Xerox Corp.
800 Phillips Rd.
M/S 111-02J
Webster NY, 14580-9701
-----Original Message-----
From: Stefan Andersson [mailto:stefan.andersson@axis.com]
Sent: Wednesday, March 24, 1999 9:46 AM
To: Zehler, Peter; ipp@pwg.org
Cc: Hastings, Tom N; anthony.porter@computer.org
Subject: RE: IPP> MOD - Use of time [and Bake Off 2 Issue 17]
On Wed, 24 Mar 1999, Zehler, Peter wrote:
> The HTTP layer requires a time stamp. Simple time protocols exist =
that
> are very light weight. A page can be made available in the embedded =
web
> server to set the current date and time. If all else fails the time =
can
be
> grabbed from incoming HTTP packets.
>=20
It is true that the HTTP layer requires a time stamp, but there is an
exception to this rule. From draft-ietf-http-v11-spec-rev-06.txt=20
14.18 Date:
3. If the server does not have a clock that can provide a
reasonable approximation of the current time, its responses MUST =20
NOT include a Date header field. In this case, the rules in
section 14.18.1 MUST be followed.
To add a page for setting the current date and time is not a good idea. =
If
a date is to be useful the client must be sure that the clock isn't
reseted between requests. The same argument can be used against =
grabbing
it from incoming HTTP packets, the server will have no way to control =
that
the incoming date/time is correct. The only way to get a date that =
would
work is to use for example NTP.
PZ> Software does not have to blindly accept a time change.
PZ> Timestamps can be adjusted based on the latest truth.
PZ> (I assume the client will ask for all the timestamps instead
PZ> of one at a time)
PZ> Across the Internet the chances of the Client and the Printer
PZ> having their time in sync is remote. A man with one watch=20
PZ> knows what time it is, a man with two is never really sure.
PZ> My point was that there are a number of methods to obtain the
PZ> current time. Each has certain advantages and disadvantages.
PZ> Let the method of obtaining the time be selected by the =
implementation=20
PZ> based on product requirements. There is a real end user advantage
PZ> to having the timestamps=20
Since the Date field probably is most useful for clients when the =
server
is spooling. I would recomend one of this alternatives:
1) Add optional attribute that gives the time in an absolute form.
2) If the server can answer in an absolute form it should do that
otherwise it will answer in an relative form as the protocol states
today. =20
PZ> Are you suggesting alternate syntaxes for the same attribute?=20
PZ> Are you suggesting the following from the issues list?
> 1.Add to the IPP/1.1 Model and Semantics document OPTIONAL job
> description attributes: .date-time-at-creation (dateTime)., .date-
> time-at-processing (dateTime)., and .date-time-at-completion
> (dateTime)..
PZ> It is statement 2 above that confuses me. Are you suggesting
PZ> the client ask for all attributes or both the absolute and relative
PZ> attributes or something else?
/Stefan
-- Stefan Andersson Software Engineer Print Server Business Unit Stefan.Andersson@axis.com AXIS Communications AB Phone: +46 46 270 19 85 Scheelev=E4gen 16 Fax: +46 46 13 61 30 S-223 70 LUND, SWEDEN http://www.axis.com