home *** CD-ROM | disk | FTP | other *** search
- Submitted-by: stephe@usenix.org (Stephen R. Walli)
-
- Peter Collinson <pc@hillside.co.uk> reports on the April 8- 12, 1992
- meeting in Dallas, TX:
-
- [Ed. - Peter is the USENIX Institutional Representative to TCOS-SS,
- the IEEE group responsible for drafting the POSIX family of standards.]
-
- Overview
-
- Theoretically, I spent most of the week in POSIX.1, the working group
- for the ``original'' system interface standard. It's still meeting
- because it has several extant projects:
-
- - POSIX.1LIS, programming language independent POSIX.1;
-
- - POSIX.16, the C binding to POSIX.1LIS;
-
- - POSIX.1a, the place where bug fixes and new features
- for POSIX.1 are being put while the language
- independence work is being done;
-
- - POSIX.18, the POSIX Environment Profile. It's a
- profile (or list of other standards) intended to
- describe something close to a complete UNIX system.
-
- I tend only to attend the work for the first of these because I also
- go to many other steering committee meetings. Here's an idea of what
- happened in the bits that I managed to get to ....
-
- The Report...
-
- The ISO standards working group on POSIX, WG15, requires that the IEEE
- POSIX working groups produce a programming language independent
- version of the existing POSIX.1 standard (ISO 9945-1). This language
- independent specification (LIS) is referred to as POSIX.1LIS.
-
- The POSIX.1 standard has been re-cast in two sections: the language
- independent specification and a C language binding (POSIX.16). The
- idea is that these two should ballot together, so that balloters can
- compare the original standard with the new pairing.
-
- It's planned now that the two standards will go to ballot on July
- 7th. This has been made possible because:
-
- - the documents are close to being ready, have been mock balloted
- and finally preened by the working group,
-
- - the Steering Committee on Conformance Testing (SCCT) has agreed
- that the documents do not need a completely new set of test
- methods written for them. They can use the already existing test
- methods for POSIX.1, contained in POSIX.3.1, which has nearly
- completed balloting.
-
- Not needing new test methods is a great concession because it avoids
- the rules that insists on test methods being available for all new
- standards before they go to ballot. In my opinion, someone will need
- to find some funding to get the new test methods written. There is no
- enthusiasm for doing this in the working group. This is also the
- consensus of the group, when asked just that question.
-
- What are test methods? That's a little hard to explain. Basically,
- they are terse English statements that assert facts about the
- standard. The idea is that these are easier to convert into programs
- that actually test the interfaces. Each assertion is classified as
- ``testable'' or ``not testable'', and whether or not it applies to
- optional behaviour. It's a little more complex than this. Look at
- POSIX.3, (IEEE Std. 1003.3-1991,) the standard for test methodologies
- for POSIX, for more information.
-
- The current document drafts are based on the ordering in 9945-1. This
- is good because sections in all the documents refer to the same
- material. If you are looking at Section 3.2.1 in 9945-1:1990, then
- the same material will be found in the same numbered section in
- POSIX.1LIS and POSIX.16.
-
- A small group of people who are close to the document: the editor (Hal
- Jesperson), the person really running the LIS project (Paul Rabin from
- the OSF) and the chair of the POSIX.1 Working Group (Donn Terry) have
- realised that this is POSITIVELY THE LAST CHANCE to change the
- ordering of the document. [Ed. - the close() and open() functions are
- in different chapters of the standard, as an example.]
-
- Donn has come up with a potential re-ordering and this will be applied
- to the new documents. I was concerned that this would make balloting
- difficult, because we lose the ability to easily cross reference. The
- idea is to print a re- ordered version of 9945-1 (without rationale)
- to act as a balloter's aid.
-
- The two new documents will also contain ``other editorial changes''.
- The adoption of the LIS has meant that the original text has been
- inspected very closely indeed, and has been found wanting in many
- places. It's often ambiguous with unclear wording. The text has been
- tightened up in these places. One of the tasks of the working group
- this week has been to examine a list of lines containing ``may'',
- ``can'', ``cannot'', ``system defined'' and some other words to ensure
- that they are all used consistently throughout the documents. Where
- ambiguities exist the wording has been repaired.
-
- Now, you may argue that this will change the sense of the document and
- it might. It will be up to the balloting group to worry about that.
- There are NO conscious changes.
-
- New functionality and real bug fixes have been held over in POSIX.1a.
- There was no discussion on this during the week, because the person
- driving that, Roy McKean from X/Open, was unable to be in Dallas.
-
- Volume-Number: Volume 28, Number 33
-
-