Requirements for a network printing environment

[Steve Smith <[e-mail]>]

print-wg

	I lead the development of a moderately complex system for
	providing output services to a moderately large networked
	customer base.
	
	We have an existing infrastructure which is overdue for
	an overhaul.  I am working on strategy plans and requirements
	for such a system.
	
	I would like to develop a model with as large of scope as
	possible and then implement whatever subset vision and
	resources will allow but "within the context" of a larger,
	more general model.
	
	The scope of the model I refer to is a
	"cooperative, distributed printing environment" which
	includes the issues of automatic brokering of services
	between environments with automatic resolution of all
	the attendant issues of accounting, quality control
	and scheduling.
	
	A subset of this model, more likely to be of interest to 

	others would be a "network printing environment" which
	is approximately what we find various systems evolving
	toward.  

	

Requirements for a network printing environment-

	Ubiquity -
	Transparency -
	Extendibility -
	Generality -
	Interoperability -
	Reliability -
	
	Printing should be taken to mean the recording of information
	on physical media.  In general it means images on paper or
	film but is not neccesarily limited to this definition.
	
	Ubiquity -
	
	The environment should reflect the relative ubiquity of
	image formats, computing environments, networking
	environments and output equipment.
	
	Transparency -
	
	Access to printing resources should be transparent.  Printing
	a file at a remote location should be as easy as printing it
	in your office.  Printing a sun raster file should be as easy
	as printing an ASCII text file.  Printing any image on a 36"
	plotter should be as easy as printing it on a laser writer.
	
	The incremental requirements to a client to use a "different"
	service from the one they are accustomed to should be only
	those directly related to the service.  A user of a remote
	printer needs to know where to pick up output or how to have
	it delivered and the expected schedule.  Someone printing an
	image designed for a 35mm slide on a color viewgraph needs to
	understand the difference in media quality.  Someone logged
	into a Unix machine from their Mac needs to understand the
	Unix environment and that lpr is required to print something.
	But no more.
	
	Extendability -
	
	The lpr suite of tools was designed literally for line
	printers but was quickly extended to include electrostatics
	and ultimately many imaging devices.  lpr is extendible but
	has been extended far past the paradigm it originally 

	represented.
	
	It should be assumed that whatever questions we ask today
	will have increasingly complex answers as they evolve and
	that we have not asked all the questions today.
	
	There are many examples in existence today that fit the
	requirement of "imaging information on physical media"
	that are not taken into account in current models.
	
	For example, the requirements of rendering 3D models into 

	3D models such as by using holography or stereolithography
	are like not to be addressed by any "practical" system, yet
	are likely to be viable technologies before the systems
	being designed to day will be replaced.
	
	The requirements of movies, multimedia, hyper-documents,
	3D, and many other areas are not critical today but may
	dominate in the future.
	
	A model that anticipates such things but doesn't try to
	address them in detail is likely to be much more viable
	than one which ignores them or solve them in detail would.
	
Generality
	Each element in the environment can be designed to "just
	meet" requirements.  It can also be designed to embrace
	a larger model, a larger view of the problem at hand.
	
	Many of the issues of extendibility are resolved naturally
	by simply designing and implementing to the most general
	model available.
	
	Spooling, queueing, scheduling, routing, accounting and 

	distribution are all issues which are addressed in other
	fields of play and the lessons learned in more mature
	arenas will most likely ultimately apply to our game as
	well.  We have enough problems of our own without ignoring
	the ones that have already been solved.  


Interoperability -

	The requirements of interoperability with other 

	installations, existing systems and systems being designed is
	also high.
	
 	It is my assumption that other installations have
	requirements that differ from ours and in some cases,
	possibly exceed ours.  It is also my assumption that if we
	could ever "union" the requirements under
	a common model, subset implementations would be interesting
	and often interoperable.

Reliability -
	Reliability is perhaps one of the greatest challenges to 

	network printing.  The nature of the work makes certain
	aspects of quality control human based and often very
	difficult.  This leads to an assumption that similar
	reliability in submission, spooling, scheduling and delivery
	is appropriate. 

	
	There are many things which can be done to minimize 

	reliability and quality issues that are not.  The requirement
	of reliability is an often undersold one in this world.
Details of my environment...
	Thousands of users from different administrative, technical,
	geographic and even "cultural" domains use various services
	offered by the PAGES (Print And Graphics Express Station)
	system here at Los Alamos.  They generate millions of images
	per month and millions of multi-image jobs per year to be
	recorded on various film and paper formats.  We even offer
	photoplot services and are developing video and considering
	CD-Rom services.  Operations runs 24 hours, 365 days.
	
	Many of these clients work from Suns, Macs, PCs, a myriad
	of Unix boxes and a moderate variety of proprietary and
	even homegrown systems.  They have varying degrees of
	connectivity ranging from standalones with floppy drives
	through modems and appletalk through ethernet to FDDI and
	Gigabit special purpose networks.
	
	Other clients work directly on the "big iron".  Connection
	Machines, Crays, IBMs, Workstation Clusters, or on remote
	machines at other labs or universities.
	
	All of these customers need access to our large volume
	resources as well as support for printing files on their
	own personal or departmental printers.
	
	Distribution of output is done via internal as well as
	external couriers, messengers and mail systems.
	
	Accounting for services is required.  No overhead $$ is
	available except possibly for R&D into new services.
	
	Classified as well as "official use" data is processed.
	
	"large" special interest communities exist.  Engineering,
	Administrative, Scientific, Facilities Mangagement, Artistic,
	and other groups drive requirements in different directions.
	
	Secretaries and Rocket Scientists have to work along side
	computer hackers.
	
***
Enough for now.  Any comments?
Steve Smith
Los Alamos National Laboratory
[e-mail]