Unix Technical Digest V1 #15

Ron Heiby (The Moderator) unix-request at cbosgd.UUCP
Wed Mar 13 14:56:56 AEST 1985


Unix Technical Digest       Wed, 13 Feb 85       Volume  1 : Issue  15

Today's Topics:
                      honeydanber uucp available
                        Micro/Systems Journal
                    number of open files (2 msgs)
                             OS INFO POST
----------------------------------------------------------------------

Date: 14 Feb 85 16:23:59 GMT
From: trb at masscomp.UUCP (Andy Tannenbaum)
Subject: honeydanber uucp available

To arms, to arms.

Sometime around the first of the year, without fanfare,
honeydanber uucp was made publicly available by AT&T.

Honeydanber uucp is the uucp system that many people inside the Bell
System use, and that many of us outside have been waiting for.  It's
named honeydanber after it's three main proponoents, Peter Honeyman,
Brian Redman, and Dave Nowitz, all one-time employees of the one-time
Bell System.

Anyway, honeydanber has been released by AT&T as:

	Basic Network Utilities Package Release 1.0

Here are the basic fees:

$5000	first CPU source license fee
$1000	sublicense fee
$250	per sublicensee binary fee

BNU is a real nice package, it's basically uucp with all the nits
preened out of it.  Everything is automatic, everything works, it has
the features you want, mail and files do not disappear, failures are
reported to the appropriate parties, efficiency hacks are all in
place, everything is configurable, it's just a win, trust me.
(BNU also has cu and ct which use the same comm routines as uucp,
finally.  It might have other goodies...)

Anyway, the prices are totally outlandish.  Let's say a fledgeling
UNIX startup has 1000 machines running uucp in the field.  That's
$250,000 we have to pay AT&T to upgrade our customer base.  It's
really not practical for us to unbundle uucp, since a functional uucp
already exists.  We can't support two differnet uucp's.

If I'm not mistaken, AT&T only charges $250 (tops) for a UNIX binary!
I (and the money people here at Masscomp) think that AT&T should
charge a much smaller per sublicense fee for BNU, if they charge one
at all.  We think that about $25/system is right.  Realize, this is an
upgrade to an existing software system which is currently distributed
without additional cost as part of all available UNIXes.

Masscomp really wants to distribute BNU, but we can't afford it, our
customers can't afford it and I don't think the story is any different
at other UNIX vendors.  It would be a shame to see honeydanber uucp
just lie there and rot because of ignorant losing pricing practices.

What to do?  Call the AT&T UNIX Software Info Hotline,
(800) 828-UNIX (828-8649) and ask about BNU and tell your account rep
that you want it but that the price is much too high.  Ask them what
you can do about it.  Write them letters.  You don't have to be a
purchasing agent for your company to call them, you can do it
yourself.

I want honeydanber to see the light of day.

The rep at AT&T who talked to me is Steve Vuksanovitch, you tell the
operator what company you work for and you might get him or someone
else.   If you want to send them snail mail, the address is:

AT&T Technologies
Box 25000
Greensboro, NC 27420

	Andy Tannenbaum   Masscomp  Westford, MA   (617) 692-6200 x274

------------------------------

Date: 13 Feb 85 03:57:38 GMT
From: abney at nbs-amrf.UUCP (Louis Abney)
Subject: Micro/Systems Journal

This sounds good.  It is being edited by Sol Libes, who edited MicroSystems
Mag for 5 years.  From what I heard (secondhand), the publishers (Ziff-Davis)
who bought it (MSM), found it more profitable to shut it down.  (The reasons
are rather bizarre to the average mortal but it apparently happens all
the time in the mag biz.  I'll cover it in another article if anyone's
interested.)

Anyway, it was a pretty good magazine, "honest" reviews and authors who
did their homework.  One thing that characterized it is that the
articles are "practical".  About the only bad thing I can say about it
is its not clear where they're going.  They were originally CP/M
oriented but they seem to slowly be phasing that out.  The last year or
so, they had lots of UNIX articles (in fact, their last issue had a
great interview with Ken Thompson and Dennis Ritchie - a real scoop).

But I think they're going to settle for the smaller system's market -
people who use MS/DOS, Xenix and code in C, Pascal, Forth, Assembler
(although they mention Lisp in the press release!?!).  The journal will
have regular columns on PC interfacing, C programming and UNIX, and
public domain software.  It says "stuff to keep every hacker up-to-date
on the ever changing micro technology"!

Micro/Systems Journal
POBox 1192
Mountainside, NJ  07092
(201) 522-9347

Its $18/yr ($24 first class).
There's a complex rate schedule with deductions if you were a MSM
subscriber (-5%) or if you subscribe before 4/15 (-%10) or if you live
overseas or ..., so you should probably call before sending money.

------------------------------

Date: 12 Feb 85 15:16:58 GMT
From: fct at raybed2.UUCP
Subject: number of open files

We are having a problem at our site in that we would like to
increase the number of files that a user can have open at one
time from 20 files to 30 files. I found where to change it in
param.h and I've remade the kernel after changing it there, but
when we reboot with the new kernel, several things don't work
correctly. If you do a 'ps', you get nothing except the header line.
Also, it shows the 'init' process having some huge amount of cpu time.
Is there any other place that needs to be changed, other than the value
in 'param.h'. Please send mail as our site has had problems getting the
news. Thank you. (We are running 4.1BSD on a vax/780)

Fred Thompson
SDF Staff

------------------------------

Date: 19 Feb 85 14:54:14 GMT
From: bill at ur-cvsvax.UUCP (Bill Vaughn)
Subject: number of open files

When you change a system header file like 'param.h', you'll have to
recompile all the system utilities which use that header file.  This is
the purpose of the 'make depend' in most of the Makefiles in /usr/src/*.
You should be able to get a list of the files you'll have to recompile
with the command 'grep param.h /usr/src/*/Makefile'.

Fantisizing, it may be useful (and sobering) to have a daemon with a full
table of dependicies send the system administrator mail reminding him of
the work ahead of him whenever he/she touches system headers.

Bill Vaughn
Univ. of Rochester, CVS

[Ed note:  I'm sure if it were as "easy" as just changing the header
file and re-compiling everything, that someone would have done it by
now.  Let's be real careful out there!  RWH.]

------------------------------

Date: 15 Feb 85 10:41:47 GMT
From: labuda at endot.UUCP (Dave Labuda)
Subject: OS INFO POST

This is a posting of information on operating system sizes.
First, I would like to thank everyone for their contributions.
Second, I would like to disclaim any connection to these numbers.
Most of the information was given to me, and I accept it as truth unless
someone claims otherwise. The UNIX info I compiled myself - it includes
all device drivers at our installation. If I missed anyone's info,
I apologize - our net connection was broken for a couple of weeks.
The number of system calls for the OS's was left out due to many flames on
the usefulness of that factor. Also, several people complained that UNIX
does not provide equal functionality as other OS's. This is somewhat true,
but I still think the chart is interesting (one can also argue that the
OS for a PDP-11 SHOULD be smaller than that for an IBM 370).
If anyone has better numbers or other systems they would like to add, feel
free to mail me the information.

Here it is, I think it shows that V7 is the clear winner :

    OS           MACHINE               SOURCE               EXECUTABLE
-------------------------------------------------------------------------

UNIX V7		  PDP-11	  ~18,000 lines (C)	       ~64K

UNIX 4.1 bsd	  VAXEN		  ~33,000 lines (C)	      ~200K

UNIX 4.2 bsd	  VAXEN		  ~75,000 lines (C)	      ~300K

VMS		  VAXEN		~2,000,000 lines (Ass)		?

DCTS		Honeywell	  ~50,000 lines (PL1)	      ~500K

TOPS-10		  DEC-10	 ~500,000 lines (Ass)	      ~500K

MTS		  IBM 370	 ~800,000 lines (Ass)		?

B6500	       Burroughs 6500	~1,000,000 lines (Ass)		?

OS/1100		Sperry 1100	~1,000,000 lines (Ass)		?

MVS		  IBM 370	>2,000,000 lines (Ass)		?

-------------------------------------------------------------------------

One last thing, the literary award of the season goes to John Muth at Sun
for comparing MVS to "kicking a dead whale down the beach".

OK - flame away folks, I'm sure no one is happy with the numbers....

						dave labuda
						decvax!cwruecmp!labuda

*There are no opinions here, so no one is represented.

*These numbers may be wrong, but at least they're not a trademark of ATT.

------------------------------

End of Unix Technical Digest
******************************
-- 
Ronald W. Heiby / ihnp4!{wnuxa!heiby|wnuxb!netnews}
AT&T Information Systems, Inc.
Lisle, IL  (CU-D21)



More information about the Mod.unix mailing list