[TUHS] PG3 or Gen3.0?

[Clem Cole via TUHS]

Different audiences/different objectives.

PG was a term and efforts directed >>inside<< the Bell System to try to get
their own different and divergent UNIX implementations together in a more
uniform manner.

SVID was much later and directed >>externally<< to folks like you and me.
It was a reaction by USG to /usr/group’s 1985 document, which was
rebirthed/rewritten in IEEE/ANSI style to become P1003.

AT&T has started their ‘consider it standard’ campaign and was very much
miffed that the ‘definition’ of there ‘thing’ was being defined outside of
AT&T and they did not control it.  In fact, there was a big effort to try
to get P1003 to take SVID directly (we pushed back, not nearly hard enough
in my eyes, but that’s a different story).  In fact, ATT never stopped, so
if you look at later editions of you can see more and more influence it had
on later documents.

To complete the thought, SPEC1170 was created when it was realized by the
members of UI (particularly the ISVs), that even with SVID and POSIX it was
still really hard to actually develop large systems oriented SW products
because while system call interface was now fairly well defined, there were
so many other ‘upper level’ interfaces [1170 of them in fact - how it got
its name] within the different UNIX implementations that described how a
system was managed and user encountered a Unix-based system, it was
extremely difficult to create just one version of your product- the test
matrix against all the different favors from SVRx, BSDs, Ultrix, Tru64,
SunOS, Solaris, HPUX, AIX370/RS/PS2, DGUX … you get the idea.


Sent from a handheld expect more typos than usual

On Thu, May 7, 2026 at 8:11 PM Larry McVoy via TUHS <tuhs at tuhs.org> wrote:

> I've never heard of PG3.  Is this something like the SVID or something
> else?
>
> On Thu, May 07, 2026 at 11:55:33PM +0000, segaloco via TUHS wrote:
> > Received a question from Thalia Archibald concerning my use of the
> > nomenclature "PG3" vs Pirzada's "Gen3.0".  I pulled together the info
> > below, seems like it might be interesting to the broader audience:
> >
> > So it's a bit fuzzy when you get into the weeds. This document[1]
> > defines the acronym "PG" as such:
> >
> > > At the highest level in the document hierarchy is the
> > > Program Document (PG) Index that lists all of the documents
> > > associated with a program generic.  A single PG-level directory
> > > occupies the corresponding position in the file system hierarchy.
> >
> > So within their source control structure you would have directories:
> >
> > pg-1c300-01 (issue 1)
> > pg-1c300-02 (issue 2)
> >
> > And then under those the "PR" directories containing the individual
> > groupings of source files, other documents, etc.
> >
> > This isn't what I originally based my use of this acronym on though,
> > rather just Program Generic (PG).
> > Typically if I do see a shorter name,
> > the system is just referred to as a "Generic", proper noun. This was
> > common practice within AT&T and did not just apply to UNIX[2][3].
> > In fact, this first round of USG UNIX (before UNIX/TS and later) was
> > very "Bell-ish" compared to later efforts. The manuals were the only
> > ones rendered to Bell System standards in that they had the OSD headers
> > with document ID, issue, date, etc. rather than just the standard
> > manpage masthead. Several of the USG documents of the time look like
> > the same typesetting package used for 3ESS generic pidents for instance.
> >
> > As I was gathering some of this info, I keep seeing PG-1C300 used where
> > the system is referred to in short.  This actually makes sense given
> > AT&T's use of the "PG" prefix for a number of different projects.
> > For instance, MERT Release 0 is given as PG-1C600[4]. COSMOS, built
> > on UNIX by WECo for a frame system in telco offices[5], has
> > documentation given as PA-6P014, possibly implying a PG-6Pxxx prefix
> > for the underlying generic. Outside of UNIX, PG-3H903 is one of the
> > 3ESS generics[6].
> >
> > After compiling all of that info, I'm inclined to say that the
> > "canonical" names in increasing length are "PG-1C300", then "Generic",
> > then "Program Generic". I would think given Pirzada's being at AT&T
> > at the time though the "Gen3.0" etc. nomenclature would also be
> > acceptable, just maybe not "period".
> >
> > - Matt G.
> >
> > P.S. As always this is just my own research, I wasn't there so I'd
> > certainly get some second opinions if available.
> >
> > [1] -
> https://www.tuhs.org/Archive/Documentation/TechReports/USG_Library/1099_Program_Generic_Control_and_Documentation_on_UNIX.pdf
> > [2] -
> https://telecomarchive.s3.us-east-2.amazonaws.com/docs/bsp-archive/233/233-154-135_I1.pdf
> > [3] -
> https://telecomarchive.s3.us-east-2.amazonaws.com/docs/bsp-archive/234/234-090-161_I1.pdf
> > [4] -
> https://www.tuhs.org/Archive/Documentation/Manuals/MERT_Release_0/Pgs
> 01-02 Title Page.pdf
> > [5] -
> https://telecomarchive.s3.us-east-2.amazonaws.com/docs/bsp-archive/SPCS/PA-6P014_I3.pdf
> > [6] -
> https://telecomarchive.s3.us-east-2.amazonaws.com/docs/bsp-archive/SPCS/PG-3H903_I10.pdf
>
> --
> ---
> Larry McVoy           Retired to fishing
> http://www.mcvoy.com/lm/boat
>