POSIX: A Case Study in a Successful Standard

1
POSIX A Case Study in a Successful Standard

• Stephen R. Walli
• Vice-president, RD
• Softway Systems, Inc.

2
Or, Why We Dont Need Radical Change in the SDO
Process

• Stephen R. Walli
• Vice-president, RD
• Softway Systems, Inc.

3
Introduction

• SDOs have been accused of being too slow to keep
  pace with the technology boom. We are told we
  need more standards faster.
• Vendor marketing practice preaches a failure to
  keep up to counter demands for standards-based
  (vendor independent) solutions.

4
Intro - 2

• Reality The standards process carried through
  the SDOs is not broken, and requires no
  revolutionary change beyond simple evolution.
• This paper looks at POSIX as an example of a
  standard caught between vendor marketing and the
  real world.

5
POSIX

• POSIX is a family of standards to support
  application portability at the source code level.
• POSIX standards are both IEEE/ANSI and ISO
  standards.
• Based on early /usr/group work (1984) it was
  sponsored by the IEEE in 1985.

6
POSIX - 2

• The first standard was ratified May 1988, and
  aligned with the ANSI C-language standard (1989).
• The work was carried forward to ISO in 1989
  (SC22/WG15) and the first IEEE/ISO aligned
  edition was approved in 1990 (ISO/IEC
  9945-11990).

7
The Working Groups

• POSIX.1 covers just the core system interface.
• By 1990, 10 IEEE working groups were sponsored
  and upwards of 300 attendees participated
  quarterly.
• By the mid-1990s, approx. 25 projects were
  sponsored across 16 working groups.

8
The Working Groups - 2

• Co-ordination committees sprang up around the
  working group efforts on related issues (Test
  Methods, Profiles, LIS, PMC).
• Then participation began to fall. As of the
  Spring 1999, quarterly meetings are attended by
  25-40, and many remaining projects are U.S. DoD
  related.

9
The Working Groups - 3

• Participation fell due to
• Work completion of the core standards (base,
  utilities, real-time, threads, TM)
• Contentious work made little progress and
  participation waned when it was clear no work was
  progressing.
• Companies cut back standards participation
  through the deepening recession.

10
The Unsung Success

• POSIX is often painted as a failure by vendors
  fighting to differentiate product
• Failing participation (POSIX cant be relevant
  anymore.)
• Lack of functionality (POSIX isnt useful
  alone.)
• POSIX takes too long (It wont be relevant if
  it doesnt hurry up.)

11
The Unsung Success - 2

• POSIX is an incredibly ambiguous term by
  itself.
• Lack of functionality and length of time are
  marketing F.U.D.

12
The Unsung Success - 3

• The IEEE working groups have produced 27
  standards, (many ratified by ISO), in 14 years.
• Key standards include
• IEEE 1003.1-1990 ISO/IEC 9945-11990
• IEEE 1003.2-1992 ISO/IEC 9945-21993
• IEEE 1003.1b-1993 (POSIX Real-time)
• IEEE 1003.1c-1993 (POSIX Threads)

13
The Unsung Success - 4

• The core POSIX efforts have been supported by a
  powerful patron (NIST) to balance against
  powerful vendors.
• The core POSIX.1 effort is ubiquitous to the
  point of invisibility. (Every UNIX system, MVS,
  VMS, MPE/iX, and Windows NT are NIST certified.

14
The Unsung Success - 5

• As well as broad implementation, it enjoys broad
  use by developers (along with ISO/ANSI C) with
  new work being done to the standard, and original
  work being ported to the standard.
• The POSIX.1,1b,1c and POSIX.2 standards are
  also central to industry consortia specifications
  in this domain.

15
The Consortia Relationship

• The relationship between the IEEE SDO and the
  UNIX industry consortia is synergistic (if
  contentious).
• The industry consortia in the UNIX standards
  space (X/Open, UNIX International, OSF, and The
  Open Group) have all been very active
  participants.

16
Consortia -2

• The IEEE also marketed to the relevant user
  groups (notably USENIX, EUUG, and Uniforum, but
  also DECUS and Share) giving them a sponsor vote.
• The consortia also were given sponsor votes
  within the process to encourage participation.

17
Consortia - 3

• The consortia in return for their participation
  ended up with a consensus based specification, in
  which they participated, and then added relevant
  material for their communities.
• In X/Opens Single UNIX Specification (1994),
  fully 80 of the SUS was POSIX.1, POSIX.2, and
  ANSI/ISO C.

18
Consortia - 4

• With the economic fall-out of the end of the
  1990s, the IEEE PASC working groups, ISO
  SC22/WG15, and The Open Group are continuing to
  work together to ensure the POSIX standards and
  TOG specs move forward together.

19
Observations

• POSIX has succeeded in the teeth of vendor
  marketing FUD for several reasons.
• A scalable, inclusive process open to all.
• Driven by existing practice and experience.
• A powerful patron to balance the vendors when
  discussions became to contentious.

20
Open, Scalable, Inclusive

• The IEEE SDO process allows any group of people
  to come together to form a standard according to
  well defined rules.
• The process scales to the number of participants
  (both dev and mgnt).
• There are processes to reap unused standards and
  withdraw sponsorship.

21
Open, Scalable, Inclusive -2

• Consortia were certainly able to participate, and
  given some privileges in the sponsorship
  committee but NOT in the balloting arena.
• Most consortia have exclusionary membership
  criteria (namely price).
• Consortia participating then adopting SDO built
  standards works.

22
Existing Practice and Experience

• Existing practice and experience is similar to
  the IETFs requirement for multiple independent
  implementations. If at least two implementation
  groups are unwilling to do the work, what value
  does it have?

23
Existing Practice and Experience - 2

• Within the POSIX efforts, EPE was a focus for
  discussions whereby existing user code could not
  be broken.
• The standards development community also has a
  strong understanding of the real-world problem
  domain - both user and implementor.

24
Patron

• When a lot of money is at stake, the vendors are
  well represented.
• A powerful patron (NIST for POSIX) also means a
  lot of money is at stake.
• It is a balance.

25
Evolution, Not Revolution

• It must not be easy to create standards in
  contentious spaces.
• Democratic government systems have two houses
  to ensure it is not easy to pass a law. The
  proper time and review is taken.
• So it should be with standards.

26
Evolution, Not Revolution - 2

• More standards faster seems to be the motto of
  those with competing standards to bring forward.
• The economics of standards encourages competition
  around stable specifications which define a
  commodity space.

27
Fin

• Questions?
• Email stephe_at_interix.com