Dennis,
For the client example you cite that wants human notification to its
submitters and receivers, i.e., email rather than ippget, then the client
can support email itself, when it gets IPPGET notification events delivered
to it. So such a client can send email to its submitter and to the intended
recipient of the job (when the intended job recipient is different from the
submitter). Such a client knows that if the Printer does any notification,
then it will do IPPGET. So such a client that wants to receive IPPGET
notifications and convert them to mail, can do so with confidence that
IPPGET will be supported by the Printer (if the Printer supports
notifications at all).
If IPPGET wasn't required for Printers that support notification, but could
support any of ippget, indp, or mailto, then such a client would have a more
complicated implementation, having to deal with whichever Delivery Method(s)
that the Printer happened to support and convert them to mail (if mailto
wasn't supported). Even if the Printer does support mailto (as well as
IPPGET), such a client might still prefer to exercise its single path of
code that converts ippget to mailto on the client, since the mail messages
that the Printer sends might not be good enough for such a client (or at
least would look different from the mail messages that the client had
converted from ippget).
I hope these practical considerations further make it clear the
implementation benefit of requiring a client that conforms to the IPP
Notification extension to support at least ippget in order to interoperate
with a Printer that conforms to the IPP Notification extension.
Tom
-----Original Message-----
From: Dennis Carney [mailto:dcarney@us.ibm.com]
Sent: Tuesday, April 16, 2002 21:17
To: ipp@pwg.org
Cc: 'Harry Lewis'; Hastings, Tom N; McDonald, Ira; Carl
Subject: RE: IPP> RE: Mandatory Delivery Method for Notifications -
Commen ts by April 15
(BTW, I have absolutely no intention of writing a client that only does
email notifications--I personally like ippget much more than email
notifications.)
Tom (and Carl), you say the hypothetical product discussed below isn't
going to have a "black mark" because it doesn't do ippget notifications,
but then you point out that customers might not buy the product if it can't
claim compliance. That is probably the ultimate black mark--lost sales.
If "ippget" and "mailto" notifications were really two different ways to do
more or less the same thing, I wouldn't mind making one mandatory. But
ippget makes sense for programs to use and mailto makes sense for humans to
use. For some products, mailto makes sense and ippget doesn't. Those
products can know that ippget is the only one guaranteed to be supported by
the printer, but if programmatic notifications aren't useful to them in the
first place, they don't get more useful just because many printers can
supply them.
So, I've said my two cent's worth (yeah, I know I probably stretched your
patience and got 4 or 5 cent's worth! ;-) on this subject. Thanks for
listening.
Dennis
"Hastings, Tom N"
<hastings@cp10.es To: Dennis
Carney/Boulder/IBM@IBMUS, ipp@pwg.org
.xerox.com> cc: Harry
Lewis/Boulder/IBM@IBMUS, "Hastings, Tom N"
Sent by:
<hastings@cp10.es.xerox.com>, "McDonald, Ira" <imcdonald@sharplabs.com>
owner-ipp@pwg.org Subject: RE: IPP> RE:
Mandatory Delivery Method for Notifications - Commen
ts by April 15
04/16/02 12:40 AM
Dennis,
Your example is a client whose implementor has decided not to conform to
the
Notification specification (when it only supports email). Don't feel bad!
The IPP Police aren't going to come and get you. Your product isn't going
to have a black mark. It just can't advertise that it conforms to the IPP
Notification specification. Its ok not to conform. The idea of client
conformance is that if a client conforms, then it can interoperate with a
conforming Printer.
Since this is a conditional conformance requirement for both client and
Printer, a Notifications conforming client and a Notifications conforming
Printer will be able to interoperate notifications.
Think of a customer who is buying a client and a Printer. If both say they
conform to the IPP Notifications, then that customer know that these two
products will be able to exchange notifications (at least IPPGET).
In you example, the customer wouldn't be mis-led into buying your client
(which doesn't claim to conform to IPP Notifications) and a Printer which
does conform to IPP Notifications, and then discover that they don't
interoperate (say, the conforming Printer doesn't support mailto, but only
IPPGET).
OK?
Tom
-----Original Message-----
From: Dennis Carney [mailto:dcarney@us.ibm.com]
Sent: Monday, April 15, 2002 18:53
To: ipp@pwg.org
Cc: 'Harry Lewis'; Hastings, Tom N; McDonald, Ira
Subject: RE: IPP> RE: Mandatory Delivery Method for Notifications -
Commen ts by April 15
So, let's say I have an IPP client that does not have a status GUI (job
submitter only); or an IPP client that implements status, but does it by
polling.
Then I see the notifications drafts and I think that a nice option to allow
my users is that if the printer supports e-mail notifications, I will allow
them to give me their email address and I will register them for email
notifications.
However, my client cannot claim to be an "IPP notifications compliant"
client because it hasn't implemented ippget. So to become compliant, I
must "implement"/"support" ippget, *even if* my client is not going to use
it? Hmmm, so since I'm not going to use it, I guess I don't have to test
my implementation very hard. But come to think of it, if I'm not going to
use it, no one can actually tell (by sniffing or any other method) whether
I have implemented it correctly (they can tell I didn't use it, but can't
tell whether my unused implementation code works or not). So I guess I can
just claim I support ippget--no one can prove me wrong!
This is what I'm trying to get at when I wonder what it means to force a
client to support something. Each client uses the interfaces necessary for
it to get its job done. Forcing it to issue ippget calls then simply throw
out the results, just so it can be compliant and have the buzzwords on its
marketing material, seems to be a mistake to me.
Dennis Carney
IBM Printing Systems
"Hastings, Tom N"
<hastings@cp10.es To: "McDonald, Ira"
<imcdonald@sharplabs.com>, Harry
.xerox.com> Lewis/Boulder/IBM@IBMUS,
"Hastings, Tom N" <hastings@cp10.es.xerox.com>
Sent by: cc: Dennis
Carney/Boulder/IBM@IBMUS, ipp@pwg.org
owner-ipp@pwg.org Subject: RE: IPP> RE:
Mandatory Delivery Method for Notifications - Commen
ts by April 15
04/15/02 12:26 PM
Harry,
I agree with Ira. Furthermore, the advantage of having the client
conformance requirements (as well as the Printer conformance requirements),
is that a client implementor knows that his/her conforming notifications
implementation will interoperate with any Printer that is also a conforming
notifications implementation.
BTW, the reason that we didn't have client conformance requirements in the
original Notifications and Subscriptions spec, is because we didn't have
agreement on any REQUIRED Delivery Method for the Printer when it supports
Notification. So now that we have agreement for the Printer, we can add
that requirement for the client (conditional on the client supporting
notifications at all).
Tom
-----Original Message-----
From: McDonald, Ira [mailto:imcdonald@sharplabs.com]
Sent: Monday, April 15, 2002 10:51
To: 'Harry Lewis'; Hastings, Tom N
Cc: Dennis Carney; ipp@pwg.org
Subject: RE: IPP> RE: Mandatory Delivery Method for Notifications -
Commen ts by April 15
Hi Harry,
To prove interoperability in protocols, the IESG always requires
conformance requirements for both Client and Server (Printer, in
our case).
But the requirement will be conditional:
If an IPP Printer supports notifications, then it MUST support
the 'ippget' in-band delivery method.
If an IPP Client supports notifications, then it MUST support
the 'ippget' in-band delivery method.
OK?
Cheers,
- Ira McDonald
High North Inc
-----Original Message-----
From: Harry Lewis [mailto:harryl@us.ibm.com]
Sent: Thursday, April 11, 2002 4:04 PM
To: Hastings, Tom N
Cc: Dennis Carney; ipp@pwg.org
Subject: RE: IPP> RE: Mandatory Delivery Method for Notifications -
Commen ts by April 15
I support mandating IPPGET for IPP notification but I believe the mandate
should apply to the printer not the client.
Architecturally, mandating support (for IPPGET) at the printer is
sufficient to support complete interoperability. It is overkill to mandate
both ends. For IPPFAX (a specific application of IPP) I agree we want to
lock down the specification for every node.
Are you are saying that someone can build a client with (ex.) mailto
notification support (only)... but they just can't claim it is an IPP
client that supports notification? If someone chooses to do this, of
course they run the risk of reduced interoperability. That is a conscious
choice. The architecture and specifications still fully support complete
interoperability.
----------------------------------------------
Harry Lewis
IBM Printing Systems
----------------------------------------------
"Hastings, Tom N" <hastings@cp10.es.xerox.com>
Sent by: owner-ipp@pwg.org
04/10/2002 06:08 PM
To: Dennis Carney/Boulder/IBM@IBMUS
cc: ipp@pwg.org
Subject: RE: IPP> RE: Mandatory Delivery Method for
Notifications - Commen ts
by April 15
Dennis,
Yes, I was adding to Carl-Uno's proposal, because in order to get
interoperability between a client and a Printer you need to have
conformance
requirements for both the client and Printer, not just the Printer. The
reason we left out client conformance in the Notification spec was because
we didn't have any Delivery Method REQUIRED for the Printer, if
Notifications were supported. But now the proposal is for the
Notification
spec to REQUIRE the Printer to support (implement) the IPPGET Delivery
Method, if the Printer supports Notification.
However, I don't think that there is the problem that you think for adding
this client requirement. My entire statement was:
2. REQUIRE that a client support the IPPGET Delivery Method, if it
supports
IPP Notifications.
The "if ..." doesn't require clients to support (implement) Notifications.
But if a client does support (implement) IPP Notifications, then it MUST
support (implement) IPPGET Delivery Method and MAY support (implement)
others. Then such a conforming client can interoperate with a conforming
Printer, including Notifications. None of these statement require the
client to actually *use* Notifications, if the user doesn't want to.
OK?
Tom
-----Original Message-----
From: Dennis Carney [mailto:dcarney@us.ibm.com]
Sent: Wednesday, April 10, 2002 08:54
To: ipp@pwg.org
Cc: Hastings, Tom N
Subject: RE: IPP> RE: Mandatory Delivery Method for Notifications -
Commen ts by April 15
Tom,
I see a way in which your comments added a new wrinkle, although I may be
mistaken. I didn't get the impression in the previous messages that we
were discussing mandating that a *client* support IPPGET if it supports
any
notification mechanisms--I read Carl's "notification implementations" as
discussing IPP servers only.
What does it mean that a client "support" a mandatory notification
mechanism? If the client has no interest in actually using that
mechanism,
it doesn't make sense to force the client to implement it anyway, then
just
not use it. Am I missing something?
Dennis Carney
IBM Printing Systems
"Hastings, Tom N"
<hastings@cp10.es To: Carl
<carl@manros.com>
.xerox.com> cc: ipp@pwg.org
Sent by: Subject: RE: IPP> RE:
Mandatory Delivery Method for Notifications - Commen ts by April 15
owner-ipp@pwg.org
04/09/02 07:32 PM
Carl-Uno,
I support the proposal to REQUIRE a Notification Delivery Method so that
interoperability between a conforming client and a conforming Printer is
enhanced for Notifications.
I also support the proposal to make IPPGET be that REQUIRED Delivery
Method
by changing the IPP Notifications and Subscriptions document (which is an
OPTIONAL IPP extension document) in the following ways:
1. REQUIRE that a Printer support the IPPGET Delivery Method, if the
Printer
supports IPP Notifications.
2. REQUIRE that a client support the IPPGET Delivery Method, if it
supports
IPP Notifications.
3. RFC 2910 already RECOMMENDs that a Printer support TLS, so saying the
same thing in the Notifications and Subscriptions document would be
redundant, but we could still do that.
Compared to our other two Delivery Methods (MAILTO and INDP), the IPPGET
Delivery Method has the following advantages:
a. it is the easiest Delivery Method to support
b. it is in-band so it doesn't create any additional firewall problems
c. it is also useful for LAN job submission (with no firewall)
d. it doesn't create any more administrative problems
e. it is REQUIRED for IPPFAX conformance.
f. and doesn't have any SPAM problems (since the job submitter is polling
and/or keeping a channel open for notification events).
The IPPGET spec also should be changed:
4. We should also change the IPPGET spec itself from its current
"RECOMMENDED" to "REQUIRED" as a Delivery Method for an IPP Printer to
support.
Tom
-----Original Message-----
From: Carl [mailto:carl@manros.com]
Sent: Saturday, March 30, 2002 13:30
To: Carl; ipp@pwg.org
Subject: IPP> RE: Mandatory Delivery Method for Notifications - Comments
by April 15
Resend, with spelling corrected etc. The earlier message slipped away
before
I had finished it.
All,
Ned Freed communicated in an earlier message to the IPP WG, that the IESG
found it unacceptable that we had not choosen ONE mandatory delivery
method
for notifications. They would also like to see that delivery method
mandate
the use of security.
As those of you who were around about two years ago remember, we could not
reach agreement about mandating any of the delivery methods.
However, in the meantime the members of the IPPFAX project in the Printer
Working Group has reached an agreement that they will require all IPPFAX
implementions to implement the 'ippget' delivery method, and it also
requires support for TLS security.
Hence, I would like to put up the following strawman proposal to the IPP
WG
members to satisfy the IESG comments:
1) Change the main Notifiction document to require that 'ippget' delivery
MUST be included for all notification implementations, but any of the
other
two methods can also be implemented as an option.
<draft-ietf-ipp-not-spec-08.txt>
2) Put that rule also into the three delivery method documents, so it is
crystal clear what the rule is.
<draft-ietf-ipp-notify-get-06.txt>
<draft-ietf-ipp-notify-mailto-04.txt>
<draft-ietf-ipp-indp-method-06.txt>
3) Further, in the 'ippget' delivery document, we specify that TLS
security
MUST be supported.
<draft-ietf-ipp-notify-get-06.txt>
If we can reach agreement on this, I will instruct the IPP editor to
implement these changes.
I would like to get your reactions back on this proposal no later than
April
15, 2002.
Carl-Uno Manros
Chair of IETF IPP WG
10701 S Eastern Ave #1117
Henderson, NV 89052, USA
Tel +1-702-617-9414
Fax +1-702-617-9417
Mob +1-310-251-7103
Email carl@manros.com
This archive was generated by hypermail 2b29 : Thu Apr 18 2002 - 13:35:43 EDT