Pete,
Thanks for this; like Ira I don't have time right now to go through all of this, although from a quick skim it looks like this fills in some of the gaps of what we defined in the past BOFs.
More good material that can be dropped into a preliminary draft of the Cloud Imaging Model and Semantics document by someone other than the current document editors...
(don't make be break out that poster again... :)
On Dec 13, 2011, at 5:33 AM, Zehler, Peter wrote:
> All,
>> I would like to see an IPP binding for Cloud printing. I see a significant advantage to an IPP based solution given the maturity of the protocol and the industry wide support. I experimented with a cloud based print service back in 2009. I based my experiments on a WS-Print binding but an IPP binding would also work. I switched the binding from WS-Print to PWG SM and included it in v1.160 of the PWG SM Schema. It is easier for me to write Schema than IPP and easier to display. It is not difficult to move between the two mappings. The issues I did not address include an environment agnostic registration and a common security model.
>> For clarity I will refer to the Cloud based Printers as Queues and the devices as Printers even though in the IPP model they are both implementation of IPP Printers. In v1.160 of the PWG Semantic Model Schema I included two files (i.e., PwgCloudPrintQueue.wsdl and PwgCloudPrintQueueMsg.xsd) that have some of the operations fleshed out a bit. (I dropped a couple things moving from WS-Print to PWG SM.)
>> I had a couple of assumptions going in.
>> I believe that Queues and Printers contain Jobs and are stateful. There is an association between the Jobs in a Queue and the Jobs being printed on a Printer. A mapping between them should be maintained on both sides and carried explicitly in the protocol. I do not believe we should be implementing “remote markers” where Jobs only reside in the Cloud. It should be possible for either side to use IPP operations to query the state of the Printer/Queue and their Jobs.
>> I believe it is a requirement that the Cloud Print model must support both Fan Out and Fan In. In other words it should be possible for a single Cloud Queue to act as a proxy for other Cloud Queues. This will allow a Printer to interact with a single Cloud Queue but still service jobs in multiple Cloud Queues. Taking this approach allows you to do things such as set different default behaviors for Queues that all map to the same Printer. (The User does not need a Job Ticket since the intent is associated with the various Queues.) This approach also allows you to enforce different capabilities for different users. Users have permission to use specific Queues. The Queues could prohibit color printing or limit the size of a Job.
> It should also be possible for one Printer to interact with a Queue and forward Jobs to other Printers. For example a Printer in an enterprises DMZ could forward jobs to internal Printers or a software only Printer could front end legacy devices (i.e., proxy).
>> The way I broke the processing of Cloud Jobs up was; Check for Work, Lock the Job, Update the Job Status, Retrieve the Document(s), and Finish the Job. I also had another group of things that include; synchronization of Printer and Job state, Firewall Traversal, Moving from Poll Driven to Event Driven, and Eventing.
>>> Check for work:
> The Printer should be event driven when processing Jobs. See below for a discussion on eventing. The printer should query the Queue for available work.
> The Operation IsThereAnyWork
> <http://www.pwg.org/mfd/navigate/PwgSmRev1-160_ServiceOperations.html#Link1036>
> is used by the Printer to ask a Queue to see if any work is available. The request contains a list of DestinationServiceUuids (i.e. printer-uuid) where the Jobs will be printed. The response will have a list of work entries. Each entry contains the SourceServiceUuid, DestinationServiceUuid (missing in schema) and a list of Job ids. I used a UUID instead of an ID but as long as the source Queue is identified either would work.
>> Lock the Job:
> Once a Printer has determined the job it will process, it sends a DeQueuePrintJob operation
> <http://www.pwg.org/mfd/navigate/PwgSmRev1-160_ServiceOperations.html#Link1035>
> to the Queue. This operates a little differently than CreateJob in that the Printer Job identifier needs to be reserved until the operation succeeds. If the operation fails or times out the identifier and any allocated resources are discarded. The request includes the identifiers for the source and destination Queue/Printer and Jobs. The response contains the number of documents in the job as well as some job information such as the job name and owner. Although not included a JobTicket should be passed in the response as well. The payload of the response should closely match a CreateJob operation. We may also want to consider passing a JobTicket by reference which IPP does not permit at this time.
>> Update the Job Status:
> Event notifications are used to keep the state of the Job, Document and Service updated. I used WS-Eventing but an IPP notification method can be substituted. See eventing discussion below.
>> Retrieve the Document:
> The Printer retrieves the documents to process using the Retrieve Document operation
> <http://www.pwg.org/mfd/navigate/PwgSmRev1-160_ServiceOperations.html#Link1038>.
> The request identified the source Service, Job and Document number. The response includes the document number and a flag to indicate if it is the last document. Other information such as a Document Ticket or the DocumentFormat can also be included. Since the Printer does not know if the document will be pushed or pulled it must be ready to accept a DocumentUri or Document Content in the response. I used MTOM but IPP already has an acceptable encoding. It is most efficient to use print by reference whenever possible.
>> Finish the Job:
> Once the Job has been printed the final update of the Job is done. Note that events are used to handle updates as the Job is being processed. The CompleteDequeuePrintJob operation
> <http://www.pwg.org/mfd/navigate/PwgSmRev1-160_ServiceOperations.html#Link1034>
> request contains the Source and destination Queue/Printer and Jobs and the details of the Destination Job. Included is The JobState, any JobStateReasons, DateTimeAtCompletion and the ImpressionsCompleted. The PrintJobReceipt can also be included.
>> Synchronize Printer/Queue state,
> Synchronize Job state,
> Moving from Polling to Event Driven
> Firewall traversal
> Eventing:
> This nice thing about outbound (i.e., Printer to Queue) eventing in a Cloud environment is that no event subscription needs to be sent. The Printer knows the Queue is interested in Printer and Job events. Either a well-known listener port can be used or the location can be part of registration. Since the Printer knows which jobs came from the Cloud it can filter the Job events. Whenever a Printer is initialized a printer state event should be sent to its registered Queues. If the content of the event is insufficient then a one way message that looks like a GetPrinterAttributes operation response could be used.
> In order to move from a poll model to an event driven model we need a mechanism to traverse the firewall. I originally used a hacked together HTTP trickle protocol. A standardized protocol such as XMPP can be used instead. The PWG can create PWG specific stanzas that map to existing IPP operations such as GetPrinterAttributes, GetJobs or GetJobAttributes. A new stanza would be needed to let the Printer know it’s time to send an IsThereAnyWork operation to a Queue.
>> Registration:
> I did not do anything with registration other than put in a place holder. The RegisterPrinter operation
> <http://www.pwg.org/mfd/navigate/PwgSmRev1-160_ServiceOperations.html#Link1037>
> request contains some Printer information such as the identifier for the Queue and the Printer. It was not clear to me if the existence of a Queue is a prerequisite or a result of the operation. The Printer information would also include the PrinterCapabilities that we recently defined in the Job Ticket work. This contains the defaults, capabilities and document generation information. I included some “Agent” information. I assumed there would be some security information exchanged so I just put a placeholder there.
>> Comments?
>>> Peter Zehler
>> Xerox Research Center Webster
> Email: Peter.Zehler at Xerox.com> Voice: (585) 265-8755
> FAX: (585) 265-7441
> US Mail: Peter Zehler
> Xerox Corp.
> 800 Phillips Rd.
> M/S 128-25E
> Webster NY, 14580-9701
>>>> --
> This message has been scanned for viruses and
> dangerous content by MailScanner, and is
> believed to be clean. _______________________________________________
> cloud mailing list
>cloud at pwg.org>https://www.pwg.org/mailman/listinfo/cloud
_________________________________________________________
Michael Sweet, Senior Printing System Engineer, PWG Chair
--
This message has been scanned for viruses and
dangerous content by MailScanner, and is
believed to be clean.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.pwg.org/pipermail/cloud/attachments/20111213/2eff9c13/attachment-0002.html>