Performance and other comments.

kiyoinc@...

Feb 22, 2000

** Reply to note from hercules-390@onelist.com 22 Feb 2000 12:14:42 -0000

I just read Mark W's paper on performance. I have a couple commments.

I don't think performance is that great an issue. While there's a lot
more that can be wrung out of the Hercules Interpretor, if you really
need performance, there are other options. Two years ago, P/370
microchannel boards were going for under $10K. I expect that these are
near give-aways now, perhaps $2,000 or so. The IBM "Partners" program
has P/390E's for $500/month lease; not bad for an ESA desktop mainframe
that clocks 7.2 MIPS.

If Hercules runs at .9 MIPS or 1.5 MIPS or only .3 MIPS, that's fine. In
1969, my first employer, the State government of Hawai'i ran everything
on a S/360 model 50, 512K, 16 2314's (about a half gig of online
storage). How many MIPS was a mod 50? .25? Herc's already got that
beat.

The real benefit of Hercules will come when we have a single
self-installing CD that will install Linux, install Hercules, DASDR
and install an MVT 21.8F IPO, and IPL 351, specify system parameters...
S TCAM, S TSO... or was if F TCAM,start=TSO

A Laserjet 4si runs at the same speed as a 1403, just over 1,100 lines
per minute.

What I'm saying is, we're *this close* to being able to give everyone a
personal S/370 mainframe like the mainframe a lot of us started work on.

After MVT, comes VM/CMS, DOS, VSE, and MVS. We'd probably want 1 MIPS to
run MVS but the hardware will provide that. (I have a meeting next week
with AMD on the future of the Athlon. Questions are SMP support and when
we can expect the 1 gHz Athlon.)

Larger issues than performance are, legally redistributable MVS DLIBs and
IPOs, VSE, VM/CMS, etc. Packaging issues, do we distribute a IPO tailored
for Hercules or DLIBs and a stage1? Does general S/390 shareware belong
on the Hercules CD? What about HASP, ASP, TSO Utilities, LSPS, IUPs
(Installed User Programs)? There's 30+ years of legacy software that
fits or doesn't.

I wonder if somewhere, someone doesn't have the PL/S clone, the Exchange,
TL/I, Speakeasy, MUSIC, Wylbur, APL, Mainstream, Alpha, and all the other
products, modifications, and subsystems that ran for a while, then
vanished.

1:56 pm

Distributing a usable system (was Re: Performance)

Jay Maynard

Feb 22, 2000

I've been giving this one some serious thought, since one goal of mine is
building a usable distribution...

On Tue, Feb 22, 2000 at 08:56:26AM -0500, kiyoinc@... wrote:
> Larger issues than performance are, legally redistributable MVS DLIBs and
> IPOs, VSE, VM/CMS, etc. Packaging issues, do we distribute a IPO tailored
> for Hercules or DLIBs and a stage1? Does general S/390 shareware belong
> on the Hercules CD? What about HASP, ASP, TSO Utilities, LSPS, IUPs
> (Installed User Programs)? There's 30+ years of legacy software that
> fits or doesn't.

We've got one significant advantge over IBM: while they had to make a
universal system that would run on a wide variety of hardware, we have the
luxury of being able to specify a particular hardware configuration pretty
closely. Because of that, I think that a Hercules CD should contain:

1) Hercules, in prebuilt, installable RPMs (at least for Intel and Alpha,
since I have both, and any others we can get our hands on).
2) Hercules source as an SRPM.
3) A pregenned, ready-to-run, as fully functional as we can make it MVT.
4) Sample JCL for anything we can think of.
5) HTML pages for building and running the system, including instructions on
how to get to the same point as the prebuilt system.
6) Other prebuilt OSes, also ready to run, as we can get them going.
7) Source code for as much as we have.

At some point, we can add MVS to the mix, but I'm undecided at this point as
to whether it should replace MVT. This will depend, to me, mainly on how
well we can get it to perform. I believe that we should be shooting for the
same subsecond TSO response time that real mainframe shops tuned their
systems for.

"Fully functional" means, to me, that we should include as much as we can,
legally. Licensing issues will be the major determinant of this, and we'll
find ourselves spending more and more time working on getting licenses for
stuff that nobody sells any more. I plan to mine the CBT tape as much as
possible for useful stuff, though I don't know how much on the 249 tape (the
earliest still available, I believe) applies to MVT.

Disk space isn't an issue at this point: I use a CD to carry things home
from the office, and Hercules, OS/360, MVS, a pregenned system, and some
other stuff takes up about 115 MB at this point. I trimmed out a lot of
stuff from Rick Fochtman's CD that isn't needed int he Hercules environment;
what's left is:

os360dlibs.tar.gz, 12 MB: the DLIBs used to build MVT and the driving
system.
os360src.tar.gz, 55 MB: All of the source code, as well as some other
reference files.
mvsdlibs.tar.gz, 27 MB: The MVS distribution libraries (less ISPF).
mvsdlbsrc.tar.gz, 6 MB: The human-readable portions of the MVS DLIBs, in
Unix text file format.
hercules-1.52.tar.gz, 300 KB
os360mvt.tar.gz, 19 MB: pregenned, ready to run MVT system.

At this point, the only reason to store things as tar.gz files is for ease
of installation...using RPM makes that even easier, with some up-front cost
in packaging effort. All of this would fit on one CD even if it were
expanded.

> I wonder if somewhere, someone doesn't have the PL/S clone, the Exchange,
> TL/I, Speakeasy, MUSIC, Wylbur, APL, Mainstream, Alpha, and all the other
> products, modifications, and subsystems that ran for a while, then
> vanished.

Good question. If we can lay hands on it legally, we should.

2:28 pm

Re: Performance and other comments.

jim_elliott@...

Feb 22, 2000

kiyoinc@... wrote:
>
> I wonder if somewhere, someone doesn't have the PL/S clone, the Exchange,
> TL/I, Speakeasy, MUSIC, Wylbur, APL, Mainstream, Alpha, and all the other
> products, modifications, and subsystems that ran for a while, then
> vanished.

APL and MUSIC are both very much alive!

Regards, Jim Elliott
Enterprise Server Sales Operations, IBM Americas
mailto:jim.elliott@...
http://www.vm.ibm.com/devpages/jelliott/

2:36 pm

Re: Performance and other comments.

Tony Harminc

Feb 22, 2000

On 22 Feb 00, at 8:56, kiyoinc@... wrote:

> I wonder if somewhere, someone doesn't have the PL/S clone, the
> Exchange, TL/I, Speakeasy, MUSIC, Wylbur, APL, Mainstream, Alpha, and
> all the other products, modifications, and subsystems that ran for a
> while, then vanished.

Most if not all of these were licensed (and not free) software. In
particular, even the earliest APL\360 (yes - that is a backslash,
predating the Paul Allen DOS backslash by many years) was a chargable
product. As for a PL/S clone, the only such clone I am aware of was
the Rand Corporation's RL/S, and that was withdrawn under extreme
pressure from IBM's lawyers, around 1976 or so.

Stanford Wylbur was free, and perhaps a very early version of MUSIC
was.

Tony H.

3:55 pm

Products.

kiyoinc@...

Feb 17, 2000

** Reply to note from hercules-390@onelist.com 23 Feb 2000 00:51:37 -0000

> Most if not all of these were licensed (and not free) software. In
> particular, even the earliest APL\360 (yes - that is a backslash,
> predating the Paul Allen DOS backslash by many years) was a chargable
> product. As for a PL/S clone, the only such clone I am aware of was
> the Rand Corporation's RL/S, and that was withdrawn under extreme
> pressure from IBM's lawyers, around 1976 or so.
>
> Stanford Wylbur was free, and perhaps a very early version of MUSIC
> was.

I think most of the well known program products and subsystems had a free
or freely available precursor. I know this was the case for APL. In
1973, while I was genning TCAM for American National Life Insurance in
Galveston Texas, our R&D subsidiery, ANCOR, was building extensions to
APL. This was some kind of DOS/360 program that John Rose modified to
co-exist with MVT; it stole IO NEW PSW.

There was a PL/S clone that floated around Washington DC in the mid
1970's and subsequently vanished. I saw an article in Datamation or some
other mag in which the author of the clone stated that while IBM claimed
that he stole the source to IBM's PL/S, he had, in fact, written it from
scratch using the description in openly available publications.

Not unlike a clean-room BIOS reverse engineering effort.

A couple years ago, I spoke with several "computer collectors, mainframe
division". I was looking for a 3330-1 drive for a Y2K source recovery
job. These hardware types were concerned about losing the history of
computing, they were trying to preserve 360/65's, Cray-1's, Apple Lisa's
and such.

While that is interesting, more of the real history is the source code
and the stories of the programmers. Looks like we have a chance to save
and run the code.

Well, enough of this. I have to switch video cards on my other machine
to get Linux running correctly on it. It has some kind of new S3 RAMDAC
and Linux misidentifies it. X displays a cursor with a rectangle
attached and the scrolling a window fills the window with hash lines.

I was going to sit on the side lines and lurk but you people are having
too much fun and my specialty in the 1970's was TCAM and TSO. In the
1980's it was C. In the 1990s, it became Rexx.... and going to meetings.

As for TCAM, I don't remember but there might be a stage1 for the canned
TSO MCP. I recall installing the 5E ICR on 21.6, in any case, the TCAM
stage1 was pretty simple, except that adding full screen TSO support was
not obvious.

I had some TSO EDIT CLISTs that customized the TCAM stage2 input. Sorry,
that was 25 years ago.

4:21 am

Re: Products.

Feb 23, 2000

Tony Harminc

On 16 Feb 00, at 23:21, kiyoinc@... wrote:

> In 1973, while I was genning TCAM for American National Life
> Insurance in Galveston Texas, our R&D subsidiery, ANCOR, was building
> extensions to APL. This was some kind of DOS/360 program that John
> Rose modified to co-exist with MVT; it stole IO NEW PSW.

By 1973 the APL\360 program product was already established. I
remember it was the first blue microfiche I had encountered - a
novelty in an era when all the IBM fiche I'd ever seen was red (or
yellow for EWS). Two companies, I.P.Sharp Associates, and Scientific
Timesharing Corp. jointly made substantial modifications to APL\360,
and sold the result as APL*PLUS, but the customer had to license the
APL\360 product as well as the mods. APL\360 came in three flavours:
DOS, MVT, and standalone (which was the DOS version, except that
after initialization it overlaid the DOS supervisor with buffers).
All three versions did PSW stealing for all their terminal I/O.

Tony H.

5:05 pm

Mini distribution

Roger Bowler

Mar 10, 2000

A week or two ago there was a thread about creating a small
self-installing Linux/Hercules/MVT CD but it fizzled out
into a discussion of the relative merits of Red Hat and
Slackware. That's a pity because it would be nice for those
people who want to get a Hercules machine up and running
quickly if they could do so without needing to be a Linux
guru.

I think that whoever volunteers to do it (Ken?) should just
go ahead and do it based on whatever distribution they feel
happiest with. For what it's worth, if I was doing it I
would probably opt for Red Hat. I know we haven't got a
stable Hercules base yet (although 1.53 must be close, as we
haven't yet pinpointed any MVT/TCAM/TSO errors yet that can
definitely be attributed to Hercules faults) but it would be
good to have a distribution ready into which we can plug a
stable Hercules when it arrives.

BTW, has anyone tried JJ's SRDA patch yet to see if it fixes
the CORE IMAGE COMPLETE and roll-delete problems??

Roger

8:57 pm