home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!not-for-mail
- From: stephe@usenix.org (Stephen R. Walli)
- Newsgroups: comp.std.unix
- Subject: Standards Update, POSIX.14: Multiprocessor Profile
- Date: 29 Dec 1992 14:24:23 -0800
- Organization: USENIX Standards Watchdog Committee
- Lines: 71
- Sender: sef@ftp.UU.NET
- Approved: sef@ftp.uucp (Moderator, Sean Eric Fagan)
- Message-ID: <1hqj6nINN9fh@ftp.UU.NET>
- Reply-To: std-unix@uunet.uu.net
- NNTP-Posting-Host: ftp.uu.net
- X-Submissions: std-unix@uunet.uu.net
-
- Submitted-by: stephe@usenix.org (Stephen R. Walli)
-
- Rick Greer <rick@ivy.isc.com> reports on the October 19-23, 1992
- meeting in Utrecht, NL:
-
- The big news in the POSIX.14 working group is that we have inherited
- the POSIX.18 draft from Donn Terry and are now responsible for seeing
- it through balloting. POSIX.18 is the Platform Environment Profile,
- more commonly known as a profile to describe the traditional
- multi-user Unix platform.
-
- Having been assured that the POSIX.18 document was ``practically ready
- for balloting,'' we traded POSIX.14's March 1993 balloting slot to
- POSIX.18. Remember that this year there are so many documents in
- ballot that a strict timetable is being used to control the potential
- administrative overload. Our document's ballot slot had been
- allocated as a purely defensive measure anyway - see below. We also
- decided to keep the balloting group open right up to the last minute,
- so those interested in paying $25.00 for the privilege of complaining
- may still do so. [Ed. - This may be raised to $50.00 in the new year!]
-
- We made one major change to the POSIX.18 draft: The C language feature
- is now required. It had been optional. Our reasoning for this was
- twofold. First, we realized that because there was no requirement
- that a given implementation provide a specific language feature,
- people could write POSIX.18 compliant applications that would not run
- on POSIX.18 compliant implementations! By requiring C at a minimum,
- vendors can guarantee portability of other languages, in particular
- FORTRAN and ADA, to all POSIX.18 compliant implementations by writing
- their runtime libraries in C.
-
- Secondly, given that POSIX.18 is supposed to codify ``classic UNIX,''
- and since classic UNIX has always included a C compiler; albeit the
- ``classic'' K&R compiler, not c89; we felt it appropriate to require C
- language support in POSIX.18.
-
- The working group also made a number of minor editorial changes to the
- document, mostly removing redundant text, which brought it down to
- less than half its original size!
-
- As for POSIX.14's real purpose, the POSIX multiprocessor profile, we
- decided not to ballot the current draft after all. We had originally
- decided to put POSIX.14 out to ballot in March in an attempt to be in
- ballot by the time the Profile Steering Committee (PSC) finalized its
- rules for ``Standard Posix Profiles.'' We reasoned that if profile
- groups that were in ballot at the time the rules were adopted were
- grandfathered in such a way as to allow them to ignore said rules,
- POSIX.14 might be the only profile to which the rules applied. This
- seemed a bit unfair.
-
- It now appears, however, that all profiles will have to follow the PSC
- rules before they can come out of ballot.
-
- So we're back to proposing new MP interfaces for POSIX.1 and POSIX.2
- that would fill various semantic gaps in MP systems that will be noted
- in the POSIX.14 draft. This includes describing parallel behavior for
- a number of common utilities (e.g, make, find, grep, xargs,) as well as
- describing special MP features of system administration functions such
- as ps(1) and times(2). We also continue to argue about processor
- binding: Can we specify enough of this in an architecture- independent
- manner to make it worthwhile.
-
- One interesting point made at the October meeting was that many of the
- participants in our working group feel that our major contribution
- will not be the MP profile, so much as our monitoring of other POSIX
- work to make sure that any new interfaces do not cause major headaches
- for MP implementations (e.g, the work that we've already done with
- respect to pthreads). With this in mind, we have proposed a new name
- for the group: POSIX.14 - the POSIX reentrancy police!
-
- Volume-Number: Volume 30, Number 8
-