SC22/WG15 N380

POSIX Conformance Test and Language Independent Specification standards

Subject: LIS and CT standardisation

After having talked with a number of Danish POSIX experts and after
a formal DS meeting I will state the Danish position as follows:

We recognize the problem with less ressources in IEEE, this is
also the case in Denmark. We also recognize that most of
the POSIX stadardisation process is done with volunteer ressources.
However, WG15 has made some decisions on the ISO POSIX process
requesting specific work in the conformance testing and LIS areas.
These requests should be followed by IEEE, as long as the WG15 decisions
are not changed. IEEE should raise the problems with WG15 and
find adequate solutions. A convenient place to do this is the next
WG15 meeting in Germany next month.

We are in favour of standardisation of CT and LIS - but we are
also in favour of having some work done. If CT and LIS specs
would slow down standardisation considerably (in the order of
multiple years), we would prefer the non-LIS and non-CT standards
to proceed on the ISO level. This is in accordance with earlier DS
statements in WG15.

We see the CT as much less important than the LIS specs.
Many other standards of SC22 do not have CT specs, eg COBOL, C.

We see the LIS specs as very important in establishing POSIX
as the ISO established OS. Many experts in ISO - eg SC2 and SC22/WG20
regard POSIX as just a C based operating system, and it is cruical
here to have other language bindings to convince them that POSIX is
generic, so they would take over POSIX technology for example for
internationalization.

The LIS specification technique should be a productivity tool
here to speed up the development of language bindings.
We do not think this is the case with the current POSIX LIS
spec as examplified by 1003.1LIS and 1003.16. Danish Standards have
several times commented that we want "thinner" language bindings
compared to 1003.16, with some ideas on how to do it. We are
prepared to do further work on this, but this must be done in WG15
as DS cannot send representatives to IEEE meetings (The DS rules
do not allow us to fund such work).

Some current ideas are that the language bindings should only be
very thin and list the syntax of the binding language API, and with
reference to the LIS sections where the functions are described.
LIS specific syntax should be avoided in the language binding, which
then could serve as a "Quick Reference Guide" to the standard.
We have drafted a C language binding in this style to 1003.1
- it should be less than 20 pages - half page size, compared to the
270 page 1003.16 document.

The other relevant language bindings should be made in this way
*before* finalizing the LIS standard, so the LIS standard could
be changed to make binding easy for the relevant languages.
We hope from DS to be able to make other draft bindings for
other languages, and currently we see the changes to the LIS
to support this as minor and straightforward.

Keld Simonsen, DK