WG15 N511


	      A Discussion Paper on POSIX Application Conformance
		       UK BSI IST/5/-/15 (Posix) Panel
				 21-Oct-94


	Executive Summary:

	This paper is, as its title suggests, an attempt to encourage
	debate on the perceived problem of POSIX application conformance
	- how to determine and to 'advertise' the fact that a particular
	application is suited to a POSIX platform.

	This paper is not a statement of position.  Most particularly it
	is not a statement of a UK position.  The UK POSIX panel has
	visited the question of application conformance a number of
	times and finds three arguments presented:

		. users - who typically want POSIX-compliance guaranteed
		  with everything they buy.

		. vendors - who recognise the usefulness of POSIX as a
		  support platform, and as a marketing lever, but don't
		  want the expense of proving conformance.

		. 'standards professionals' (for want of a better
		  description) - who believe their existing efforts are
		  largely ignored by both users and vendors.


	The User's Problem:

	. The number of operating systems platforms purchased over any
	  given period of time is dwarfed by the number of applications
	  purchased to run on them.

	. The more widely accepted a particular operating system
	  platform, the richer the set of available applications will be
	  to run on it.

	These truisms are so obvious they are readily overlooked.  POSIX
	is the basis of a widely accepted (family of) operating system(s).
	Users have specified requirements for POSIX systems, have found
	them and bought them.  They have 'bought' the concept of future-
	proofing their investment in platforms and are looking for the
	applications which they can carry forward from the rich
	selection on offer  ...then they become confused:  they cannot
	identify POSIX-conformant applications from all the others.

	This view of POSIX from the user's angle is well put by Robert
	Smith, in his paper "The English NHS Requirement for Application
	Conformance Testing", (circulated by email as SC22WG15.363 and
	attached).  The demand for POSIX applications conformance is
	largely user-driven, though it gains some support from the
	'standards professionals' group.

	To summarise (accurately, I hope) the user's argument, WG15's
	effort in standardising POSIX will be still-born if it fails to
	address the user's need for identifying and 'classifying' POSIX
	applications.


	The Vendor's Problem:

	. There are solid business reasons for writing applications to
	  run on a widely accepted operating system platform (or API).

	. There is keen competition between functionally-similar
	  applications on the same platform:  costs must be minimised to
	  enhance the marketability of the vendor's own applications.

	POSIX platforms are popular; they offer a ready market for
	applications, but the cost of proving conformance could render
	the product uncompetitive.  Equally, an application which uses
	only the POSIX family of APIs may be unattractively slower than
	a similar application which uses 'extensions'.

	Vendors argue that users are ultimately fickle beings: they
	claim that portability is their main concern, but will
	ultimately make a choice of application based on cost and
	efficiency - the extra cost burden of proving conformance will
	ultimately weigh against them.


	The Standards Professional's Problem:

	. A vast amount of effort has been and continues to be expended
	  on conformance testing and certification for a wide variety of
	  standards.

	. Even when those standards are in place, users fail to adopt
	  policies which encompass those standards and thus move them
	  towards conformance of applications.

	The 'Standards Professional' believes he is producing the
	standards that the task demands, but on the one hand vendors
	tell him that the cost of conformance testing is prohibitive,
	(especially for small, cheap applications), and on the other
	hand users fail to demand proof of adherance to the standards
	that exist.



	A Suggestion:

	A principle of engineering is that a triangular structure is
	inherently strong.  The three-cornered stand-off described above
	is equally difficult to break; but common factors appear to be
	the cost and the difficulty of proving (vendors) and requiring
	(users) conformance.

	We need a cheap and easy solution.


	ISO/IEC 9945-1:1990 outlines (in Section 1.3.2) a set of four
	'levels' of application conformance to the standard.  The UK has
	proposed an additional level of 'Rigorous Conformance', (WG15 AI
	9405-04), which is under consideration.

	One solution may be to register a set of 'trade-names'
	corresponding to these levels and to allow vendors to use these
	as 'badges' on their products.	There would not necessarily be
	any requirement for proof of conformance to these levels,
	although first or third party testing would be encouraged (by
	the users).

	So how is the scheme policed, with no proof of conformance
	required?  Peer vendor- and user-pressure may be sufficient
	here, in that where a user buys a product which fails to live up
	to the claims of the 'badge', he can approach the vendor and ask
	for some explanation of the shortcoming.  If the explanation is
	inadequate the user can approach the 'trade-name' holder, who
	can, if he is not satisfied, require the removal of the 'badge'
	from the product.  The user may then have a reasonable claim
	against the vendor, certainly in the UK, for selling a product
	unsuitable for the purpose.



	All comments, criticisms, suggestions and debate are welcome.  A
	set of the email exchanges forming the background to this paper
	are available as UK P261.