The History of IDMS
--- Compiled from IDMS-L and Above the Line - Edited and html'd by IDMS/SQL News
This treatise first appeared in IDMS-L in June
1998 and Aug 1998. Several 'old timers' contributed thru IDMS-L.
Later Chris Hoelscher published an edited version in 1999.
The Origins of IDMS by Peter Karasz
In the EARLY 60's, B.F. Goodrich Chemical Corporation (BFGCC) acquired a
GE-200 system in an attempt to automate more of its data processing
functions. Systems were developed using GECOM, a COBOL like language of
General Electric for the GE-200. The early compiler had a number of
problems and some of the fellows at BFGCC had provided considerable insight
in helping GE debug the software, and over time a good working relationship
developed. When IDS became available, Dick Schubert (BFGCC IS staff)
realized that this could be a powerful tool in developing on-line systems.
IDS, Integrated Data Store, was designed by Charlie Bachman for General
Electric to run on their GE-635 Mainframe under their GECOS operating
system. IDS for the GE-200 was a primitive data base manager.
It allowed a single file only for the database and had no tools for table generation so they
had to be hand coded in assembler. Mistakes in coding the tables combined
with bugs in the IDS code itself were a rich source of fatal errors that
caused the system to crash, several times a day. BFGCC was not able to
bring up a production system with GE's version of IDS due to its
unreliability.
Eventually, BFGCC offered to help debug the software if GE
would have no objection. B.F. Goodrich bought the rights and converted it to run on the
IBM family of mainframes and renamed it IDMS in the process.IDS for the GE-200 was written in the language of
the General Assembly Program (GAP). This assembler had no DSECTS, no macros
and no cross-reference. In addition, there were no EXTRN type addresses or
a linkage editor to resolve them. The GE-200 had only a loader. IDS and its
tables had to be assembled together to create the load module, which meant
a big re-assembly taking about 30 minutes, each time you changed either IDS
or the tables. Programs communicated with each other through fixed memory
locations, assigned in advance and hard coded into the programs that were
to use them. Jim Gilliam (of the IS staff) and his team cut the IDS
assembly listing apart and studied it subroutine by subroutine until they
gained an understanding of the logic used. They also realized that using
second generation assembler for implementation of complex software was a
loser's game. A new tool was needed.
GE had a software implementation language called WIZOR. The GE FORTRAN II
compiler was written in WIZOR as well as the WIZOR compiler itself. Which
made WIZOR an easily modifiable language and a new enhanced dialect, G-WIZ,
was soon born. Gilliam and others re-wrote IDS into
G-WIZ and so created a dependable version. A copy of the new software was
returned to GE as part of the agreement. BFGCC continued to work with its
version of IDS and in 1969 it was modified to handle two database files. By
this time, BFGCC was now running a (sort of) on-line
(more correctly remote batch) order processing system in operation on a
GE-265 configuration (GE235 CPU with Datanet-30 network controller) using
their modified version of IDS for the GE-200 series machines.
IDMS name makes its debut
At this point, BFG instituted a corporate policy dictating that all BFG
data centers will select IBM equipment (or else) when upgrading to third
generation computers. Since there were no IDS type DBMSs available on IBM
mainframes, BFGCC management, after a feasibility study, authorized the
development of IDMS. BFGCC was a member of CODASYL and as such, had an
advance copy of the CODASYL Data Base Task Group's (DBTG) April 1971
specifications. Also, Schubert, who represented BFGCC in CODASYL, was an
active member of the DBTG and thoroughly knowledgeable regarding the
proposed specifications. IDMS was developed using a subset of the DBTG's
specifications.
IDMS was designed and written at the Cleveland data center of The BFGCC.
Most of the work took place from late 1969 to mid 1971. The project team
consisted of five programmers under the management of Richard F. (Dick)
Schubert. The programmers and their contributions were as follows: Vaughn
Austin, IDMSCALC & IDMSINIT; Ken Cunningham, IDMSDMLC; Jim Gilliam,
IDMSDBMS, IDMSIDMS, and the overall design of the IDMS system; Pete Karasz,
IDMSDUMP, IDMSPFIX, IDMSRBCK, IDMSREST, IDMSRFWD, IDMSUBSC, and ISL
language compiler; and Ron Phillips, IDMSCHEM.
It is without doubt that the work with IDS gave us the confidence to
tackle development of IDMS. It is important at this juncture to stop and
point out the facts that the only IDS source ever seen by us was the
earliest version for the GE-200 and that no code from IDS has survived into
IDMS. We did, however, retain the concepts of file navigation based on
tables and the owner-member record relationship. Multi-file databases
(DMCL), areas, schemas (DDL) and subschemas are all CODASYL concepts
implemented for the first time in IDMS. The IDD was our original
contribution and it was not part of the DBTG's specifications until years
later. IDS for the GE-200 had none of these features and for that reason
alone it could not serve as the basis for IDMS. IDMS was all new design,
all new code. Later versions of IDS for the GE-400 and GE-600 series
machines were CODASYL compliant implementations but they came years after
IDMS.
Just what WAS ISL? Although the IBM assembler was much more sophisticated
than GAP, our aversion to assembler for large-scale software implementation
has remained with us. A language like WIZOR was needed for the IBM. Recall
that this was 1969. "C" is not yet a gleam in the eyes of Kernighan and
Ritchie. IBM is already using PL/S for writing parts of MVT and other
software, but they are keeping it proprietary. We had to roll our own. The
implementation language was developed with the required IBM/360 features
using GE-200 WIZOR 5 as the starting point which gave the programmers a
familiar syntax to work with. Coding of IDMS has begun months before there
was a compiler to grind the code. The code generator part of the compiler
was chopped away and new code written to produce 360-assembler source code.
We ran test compiles of small routines and captured the output on 7-track
tape, which the GE-200 and the IBM/360 could both read. The code was
submitted to the IBM assembler and if errors were found, we adjusted the
compiler and tried again. Eventually we got error free assemblies. At this
point we could compile the compiler and assemble it on the 360.
Intermediate Systems Language (ISL) was ready for business. For the IDMS
implementation we had to make do with a primitive one-pass version, as
there was no time for unessential improvements. It did the job. Later, a
multi pass optimizing compiler for ISL was written to produce code for, and
take advantage of all addressing modes of, the PDP-11/45. A highly
optimizing version for the IBM-370 followed in 1975. IDMS and all its
utilities were originally written in ISL except IDMSDMLC, IDMSCALC and
IDMSCHEM.
At this point, IDMS was ready for production. The initial set of data base
tables was hand coded, first in plain assembler, later using macros. The
macros were also utilized in creating the first IDD in early 1971 so the
schema and subschema compilers had a dictionary to work from. Once those
processors were developed, the hand coding of subschema tables came to a
welcome end. While the development of the IDMS software was in progress,
application programmers were busy converting the components of the order
entry system from GECOM to COBOL. The first production on the IBM 370/155
using IDMS was PRESTO, an accounts receivable system. TOPSY, the order
entry system followed shortly and many others later. IDMS at this point =
was a local mode only system.
How did IDMS come to exist outside of BFGCC? BFGCC realized that IDMS was
marketable software that would find a ready audience among those who, like
BFGCC, were early users of IDS but were now confronted with the need to go
to an IBM mainframe. A proposal was submitted to BFG management to
establish an independent business unit, the Information Systems Division,
which would market IDMS and other software yet to be developed and also
support BF Goodrich with its corporate data processing needs. Coding pads
with the ISD logo were printed and the first edition of the IDMS User's
Reference manual bore the ISD inscription as well. Promotional materials
like IDMS ashtrays with miniature Lifesaver Radial tires around them were
given away to prospective customers. BFGCC began selling copies of IDMS,
first to a local company, then to
others. The first five customers of IDMS were: ACME Cleveland Co., Abbott
Laboratories, General Motors, RCA, and Sperry Rand (UNIVAC). Two
more prospective customers, Boeing Computer Services and Western Electric
were trying out IDMS but we never got to sign them because the
roof fell in on our hopes of becoming an independent division within BF
Goodrich.
Enter Cullinane Corporation
In 1969 Ben Heineman, CEO of Northwest Industries, has launched a hostile
takeover attempt on BF Goodrich which failed but left Goodrich in a
vulnerable financial position. Corporate management was busy keeping the
Company afloat and had no time or inclination to consider a new business
venture, especially one that had an intangible product that was totally
unfamiliar to them. At this point it is not clear whether we were turned
down in Akron or if BFGCC management scrapped the idea without asking.
Since BFGCC had extensive experience in licensing many of its proprietary
chemical processes, we were told to take advantage of that expertise and
license the product for marketing to another company using some sort of a
royalty arrangement. After some research, Jim Gilliam recommended John
Cullinane and his Cullinane Corporation. Big John named Tom Meurer project
manager, who hired Ron McKinney and Bob Goldman from AT&T Long Lines, and
Dave Thole, a freshly minted graduate from the University of Dayton. Don
Kraska from BFG Corporate, with more IBM experience than any of us, was on
loan to BFGCC to help with the creation of the IDMS installation tape and
to help with the DMCL implementation. Shortly after that Cullinane made
their first installation of IDMS at a site we already had lined up. The
rest, as they say, is history.
Since then, in 1986 BFG began its transformation from a rubber and
chemicals company
to one primarily involved in aircraft support systems and maintenance. The
Tire Division was combined with Uniroyal's as the Uniroyal-Goodrich Tire
Company. Within a year or so BFG sold its interest in the company and was
out of the business it was best known for. Next came BFGCC. The Company was
busted into three divisions. One sold outright, one kept as the Specialty
Chemicals Division, and one spun off as an independent entity called the
Geon Company. Specialty Chemicals switched to an AS400 and left IDMS around
1995. Geon went to a SAP based client server system on Digital (excuse me,
Compaq) Alpha processors and completed the move of all production from IDMS
in February of 1997. The IBM mainframe was removed in April of 1998. And
the original development team? Vaughn Austin left BFG in 1995 and is in
Technical Systems Support at Alltel's Twinsburg, OH Data Center. Ken
Cunningham left BFG in the 80's, worked at various companies in the
Cleveland area and is now retired. Jim Gilliam, the "father of IDMS" still
works at Geon, has been an Oracle DBA these last three years and was
expected to retire at the end of 1998. Pete Karasz served out his time at
Geon and took an early retirement offer, have been out since January of
1998. Ron Phillips remained with the Uniroyal-Goodrich Tire Company until
Michelin bought it and moved everyone to Greer, SC. Ron passed away when he
suffered a heart attack while clearing his drive after a freak snowstorm.
Dick Schubert retired from Geon in 1995.
The story supplied by the original Developer Peter Karasz ends here. We continue the story:
IDMS History - Part II
Cullinane Database Systems developed the system further. The main enhancement was the addition
of an integrated online component - IDMSDC. In addition to its usage as a database server to be
used with IBM's CICS, now IDMS could support a complete DB/DC environment of its own. Rick Nini
(R) and Don Heitzman (H) led the DC development team , who also contributed to the 'RH' in RHDC
prefix of IDMS modules. With the addition of ADS/OnLine 4GL IDMS was the leader in the database
market. The dictionary - IDD - became a complete online dictionary - the only active dictionary in the market that time. A datamation report dated 1984 said that 4 out of 5 new IBM customers opted for IDMS.
But Cullinet did a marketing blunder with the Release 10.0 of the product. That was advertising the product as relational! Recall the big IDMS/R advertisements?
ASF component was the 'relational' part! ASF was built using LRF which itself was not bad as long as one restricted the usage to retrieval only (like VIEWs in relational systems). One could make a table and a builtin map and dialog automatic. But there it stopped. Database is more than a simple standalone table. Using ASF one could join two tables. But there was no way to connect a group of tables to a dailog or a COBOL program. Besides the all important 'SQL' was missing.
In short, there was no real R in IDMS/R. Eventhough the other components in the group, ICMS and Goldengate were good for 1984 (Goldengate/ICMS provided the first integrated PC access to mainframe data!), the R failure overshadowed evreything! Good or bad is a different thing - one cannot claim one's product to be relational without having a relational product and without endorsing the model itslef! This eventually dragged IDMS into the discussion with Dr Codd (the infamous 12 Rules). 12 rules were not the Bible of Database Theory, but it served its purpose - broadcasting and proving IDMS/R as conforming to none of the 12 Rules!
[In reality, there was no point in including IDMS into the 12 Rules discussion, beacuse IDMS was not a relational system, then. If Cullinet hadn't claimed IDMS to be relational, then Codd could not have included the product in his comparison (eg: IMS was not included in the comparison)]
1987 Userweek had a blueprint for IDMS release 11.0 with full SQL support. SQL Option was supposed to be based on the SQL Engine of VAX-DB with its optimizer written in C. This was the right direction to go for IDMS. But Cullinet management headed by John Landry & Co had other ideas. They wanted to focus on Expert Systems. IDMS was sidelined and the next Userweek in 1988, there was no talk of Release 11! What a blunder. Many who worked hard to get Release 11 left the firm. VAX Magic did not bring enough revenue. Then John Cullinane came back for a brief period and talked about Enterprise DB for MVS! That is SQL Option was to be a separate database and not part of IDMS! Well, the drama did not go further. In the summer of 1989 Cullinet ran out of money and fallen in the hands of CA. By October 1989 the takeover was complete. CA got a quality database with its 3000 customer base. With Cullinet acquisition, CA crossed the 1 billion dollar mark in revenue.
Though CA released Release 12.0 with full SQL in 1992-93, there was no attempt to popularize SQL. Either CA did not know what they had (in relation to other relational systems) or IDMS's own faithful CODASYL people did not trust SQL. Whatever be the reasons net result is the same: SQL option of IDMS is scarcely used and many industry analysts do not consider (quite wrongly in fact) IDMS as a relational database.
When DB2 V 4.x was released, the product had just come of age! Here is one of the features 'proudly' introduced in V 4: the ability to use 'as' clause in SELECT! Version 4 This feature was in IDMS R12.0 and even in ADS+/PC SQL! "As" caluse was also supported in OLQ/SQL of 10.2. It's hard to believe IDMS gave up against such an incomplete product so early and so easily!
If DB2 had been the 'IDMS' killer, then by now DB2 should have monopolized the database market, which is not the case today. In fact there were no 'IDMS killers' other than Cullinet's foolish decisions and later CA's 'lack of any decisions'! There was a vacuum in the db market which was filled by products like Oracle, Sybase and Informix. Even here (UNIX market) the better product was Ingres which failed. Oracle with its agressive marketing forged ahead. But is it marketing alone? Not at all. Oracle put their technicians at the site as consultants. It served two purposes: One it made more revenue than the product sale on the mini machines. Secondly, Oracle was able to support the sub-standard database with their own people at the site and mask the errors and shortcomings from the real users. Later years proved that this strategy worked very well in an ever changing "Client/Server/Unix World"!
Anybody remember where Rini and Heitzman went after leaving Cullinane/Cullinet?
Well, they formed BST (Business System Technology) which first came out with RTE ...which was bought by Cullinet and ultimately became PERFMON for IDMS. Later BST
built a product called Endevor which was eventually sold to (guess who?) CA,
and is now back in IDMS world.
How about Flip ... who left Cullinet, started DBMS (DMLO, etc)..
Flip's original partner, Ray Nawara, acquired controlling interest in DBMS with the help of Shamrock Holdings (Roy Disney, Walt's brother) so effectively DBMS became a Mickey Mouse organization. DBMS sold the education and consulting to Keane Inc. in 1989 and the products
were sold to CA in 1990. Flip then started Platinum Technology, which became a successful DB2 vendor and now in the CA basket!
ADS+/PC gave the correct results where IDMS 14.0 was totally wrong. How can one trust such a database which does not understand one of the most elementary SQL clauses? This is worse than getting an error code.
Get your own Free Home Page