Usenet 1994 January/usenetsourcesnewsgroupsinfomagicjanuary1994.iso/sources/std_unix/volume.19/text0120.txt

Internet Message Format | 1990-05-17 | 6.4 KB

From: [e-mail] (Jeff Haemer)

As editor of the USENIX Standards Watchdog Committee
reports, being bland, uncontroversial, diplomatic, or
politically correct aren't parts of my job.  Being accurate
is.  Nawaf Bitar and John Gertwagen cornered me at Snowbird
and convinced me that pieces of the section on P1004 in my
last editorial were just plain wrong.  I'll try to correct
that.

Who writes the editorials?

But before I fix that problem, I have to fix a more
important one.  John Gertwagen and Rick Greer have taken
heat for my error.  That's wrong.  Both have written fine,
useful reports for USENIX in the past about P1003.4; they'll
continue to do so in the future.  For example, in my
editorial I asked if anyone knew about an invitation-only
strategy meeting of two, large organizations to discuss
threads, rumored to be taking place before the Utah POSIX
meeting.  John's answer was, ``Actually, lots of people know
about OSF/UI coordination on threads.  They've been up-front
about cooperating and have every right to do so.  I think
it's a step in the right direction, and a sign of the
positive contribution the POSIX effort can make.  I'll say
more in my next Watchdog report.'' (John's always a little
verbose. :-) Rick's Snowbird report is already in my ``in''
box.

The contents of such quarterly standards reports are the
opinions of their authors.  Sometimes I insert editorial
asides, clarifying or disagreeing with what one of the
watchdogs says in his or her report.  When I do, I bracket
my comments, and label them clearly [Editor: Like this].
[Note that that is not the same as comments from the
moderator of the newsgroup comp.std.unix, which are marked
like this one. -mod] [And that is not the same as comments
from the publisher (even though the publisher and the
moderator both happen to be me, John Quarterman). -pub]

I also write a quarterly editorial --  my opinions on the
state of various standards efforts.  I form my opinions by
going to standards meetings, talking to and corresponding
with participants, and editing the Watchdog reports.  If you
don't like my editorials, talk to me directly or write a
letter to the editor.  My publisher, John Quarterman, will
print it.  [You can also submit something directly to the
newsgroup, comp.std.unix (perhaps by mailing to std-
[e-mail]), or to the editor of ;login:
<[e-mail]>, or you can complain to me
<[e-mail]> as publisher.  In addition to criticism,
you can post your own reports, either as a Watchdog Report
or directly to the newsgroup.  -pub] If I get something very


                           - 2 -

wrong, I'll print a retraction, like this one.  But the
opinions in editorials are the editor's.  Mine.  If you
don't like them, don't go after someone else, come after me.

Report editor eats crow

I got at least three things wrong in my editorial.  [As
already noted in a previous article, the publisher review
process (or the lack thereof for Jeff's summary article) was
also at fault for the recent confusion.  -pub] Rather than
re-state what's not true, I'll say what I should have said:

   - The ``pthreads group'' is part of the real-time group,
     not a separate entity.

   - Many people drawing up the threads proposal (.4a)
     prefer having the threads work separate from the
     current, real-time ballot.

   - There was a strong, block vote against the current
     real-time proposal (.4).  Many, but not all, of the
     initial authors of the pthreads proposal joined that
     block vote.

Next, the details.

The pthreads subgroup

There isn't one.

The .4 committee works by sending action items to small
groups for study, discussion, and recommendations; such
small groups are often responsible for one chapter of the
document.  At the urging of some ``user'' members of the
real-time group, an independent group put together a threads
proposal and presented it to the group.  That proposal was
accepted as a work item, and included as a chapter in the
P1003.4 draft.  Though the proposal (pthreads) was created
outside of the plenary meetings, when it was accepted as a
work item of .4, it became the responsibility of the entire
working group.  Small groups are not standing committees.
Over a period of a year, only one or two people may
consistently attend any particular small group.

At the last few meetings, as the rest of .4 neared ballot,
the group decided that pthreads was not sufficiently mature
to go to ballot.  it was included, as an appendix, for
comment, and the group sought and got a new PAR (.4a) for
continued work on it.  A rather large number of people whose
primary interest is in threads gravitated to the threads
``small'' group in these recent meetings, which may have


                           - 3 -

given the impression of a distinct ``subgroup.'' The
original authors of the pthreads proposal formed the core of
this group.

The separation of pthreads (.4a) from .4

I've heard two views expressed on this.  Some .4 members
think that the pthreads proposal is farther along than the
remainder of the .4 work.  These people favor a separation
to keep other parts of the real-time work from holding up
approval of a threads package.  They believe that pthreads
could come to ballot and be approved before consensus is
reached on the parts of .4 that are already in ballot.

The ``official'' story is that, Bill Corwin, the chair of
.4, suggested a separation so .4 could go to ballot while
the working group built understanding and consensus on
threads.  This was voted by the working group in Milpitas
(Nov. 89) and was the first step toward a separate PAR.

We'll see.  Perhaps the answer lies somewhere in the middle,
and the two documents will merge again before they're
through.

.4: Just say no.

Many members of the balloting group for P1003.4 don't care
for the current proposal.  Five of these drafted a common
reference ballot (CRB), ultimately filed by Mike Karels, of
CSRG, which many other .4 participants endorsed.  All of the
people drafting the ballot except Mike had been closely
involved with pthreads, but their objections were the same
as other signers of the CRB: excessive innovation and poor
design.  (Here, again, there's difference of opinion: this
time on how strong and deep the resistance to the .4
proposal is.  I think the fight will be bloody.) There seems
to be enough interest in the CRB that I'll post it
separately.

In summary

I hope I've cleared things up.  As they say in the
newspapers, ``We regret the error.''


Volume-Number: Volume 19, Number 123