Have we a Human Ecosystem?

Have we a Human Ecosystem?
James Davenport1
masjhd@bath.ac.uk
Departments of Computer Science and Mathematical Sciences
University of Bath
1
Many thanks to Capgemini staff and members of SC2
[ABB+
16],
H2020-FETOPEN-2015-CSA 712689
James Davenportmasjhd@bath.ac.uk Human Ecosystem? 1 / 17

COBOL

Why am I talking about COBOL here?

Because in 2020 the USA ran out of COBOL programmers

Because in 2020 the USA ran out of COBOL programmers (as well
as masks, ventilators etc.).

1 New Jersey Gov. Phil Murphy has issued a call for volunteers
who know how to code in the 60-year-old COBOL. “Literally,
we have systems that are 40-plus-years-old”.

2 Connecticut has also admitted that it’s struggling to process
the large volume of unemployment claims with its“40-year-old
system comprised of a COBOL mainframe and . . . ”

3 In 2008, then-California Gov. Schwarzenegger wanted to issue
minimum-wage checks to 200,000 state workers, but the
state’s controller said it would need at least six months to
reprogram California’s COBOL-based payroll system to issue
them.

them.
4 A 2018 report by the inspector general for the Social Security
Administration found that the administration maintained more
than “60 million lines of COBOL” (probably not unique lines).

them.
4 A 2018 report by the inspector general for the Social Security
Administration found that the administration maintained more
than “60 million lines of COBOL” (probably not unique lines).
5 A friend got hired back at “>$1000/day” for COBOL maint.

COBOL (continued)
Over half the states found their COBOL systems couldn’t cope
with the volume of claims. [She20]

COBOL (continued)
“IBM is addressing the call for skilled COBOL coders by working in
tandem with the Linux Foundation’s Open Mainframe Project to
create three new initiatives to address the immediate need:

COBOL (continued)
1 Calling all COBOL Programmers Forum. A new forum
where developers and programmers who would like to
volunteer or are available for hire can post their profiles and
credentials.

COBOL (continued)
credentials.
2 COBOL Technical Forum. A new resource for being actively
monitored by experienced COBOL programmers providing free
advice and expertise.

COBOL (continued)
credentials.
2 COBOL Technical Forum. A new resource for being actively
monitored by experienced COBOL programmers providing free
advice and expertise.
3 Open Source COBOL Training. A brand new open source
course designed to teach COBOL to beginners and refresh
experienced professionals. IBM partnered with the clients and
universities to develop this in-depth course which is available
for free to anyone”.

But we shouldn’t mock the Americans

[Fin19] reports a local fiasco.
Ulster Bank’s ageing systems make it impossible for the
lender to routinely charge large corporate customers in the
Republic of Ireland for deposits that run into millions of
euros, according to two people briefed on the issue.
“You wouldn’t believe how hard it is to change”.
Ulster Bank has implemented a manual workaround that
allows it to levy an ad hoc charge on a small number of
the largest depositors.

COBOL integers are decimal and unsigned by default

COBOL integers are decimal and unsigned by default, and a signed
integer (PIC S99V99) takes more space than an unsigned.

COBOL integers are decimal and unsigned by default, and a signed
integer (PIC S99V99) takes more space than an unsigned.
Remember there’s no #include.
Contrary to [Tho19], this isn’t “programmers being unnecessarily
specific”: I’d have written the same code in 1975.

COBOL: Why so difficult

After all, similarly-aged languages like FORTRAN don’t have this
problem.

problem.
1 The FORTRAN ones tend to be in active development (e.g.
Met Office).

problem.
Met Office).
2 COBOL has no equivalent of #include (and a preprocessor
providing it was my firm’s most successful product in 1975),
which gives a technical debt problem.

problem.
Met Office).
3 COBOL has no equivalent of #define, so instead of
#define BATCHSIZE 50000
you have 50000 (and 50001, 49999 etc.) throughout the code.

problem.
Met Office).
4 Y2K people wrote lots of ad hoc tools

problem.
Met Office).
4 Y2K people wrote lots of ad hoc tools
Who knows what my successor in 40 years time will write
about SPARK/cvc-5/Isabelle/. . . , or even about C?

Technology is not the answer
We’ve tried to migrate using tools to port from mainframe
code to Java. But you end up with Java code which looks
like COBOL, so your Java programmes can’t understand
the code and your mainframe programmers can’t change
it. [Sum22]

Technology is not the answer
We’ve tried to migrate using tools to port from mainframe
code to Java. But you end up with Java code which looks
like COBOL, so your Java programmes can’t understand
the code and your mainframe programmers can’t change
it. [Sum22]
The problem is the lack of a human ecosystem to understand the
technology.

And it’s not just the language
A friend (aged ∼72) works for a major bank, maintaining
mission-critical systems.

40 year old system: finally decommissioned.

30 year old system: now runs on Linux hardware, so the
Board aren’t worried any more.

20 year old system: I can’t get anyone interested in updating
this.

this.
Though I’m the only person who knows how it works,
and there’s no succession plan.

this.
The fundamental problem isn’t the technology, it’s the
management attitude to software, and the lack of “preventative
maintenance”

this.
maintenance” (which would keep the maintenance team current).

this.
maintenance” (which would keep the maintenance team current).
Also really poor human succession planning.

Italian Railways [Cim22]

Italian railways are migrating away from Relay-based Railway
Interlocking Systems (RRIS).

1 The factual documentation is on manually drawn paper
diagrams (in a mixed state of preservation).

2 The intellectual documentation (“why” etc.) is in informal
documents and engineers’ heads.

“Moreover, the work on these documents requires a strong
legacy domain knowledge which is vanishing in these days”
[ABC+20]

[ABC+20]
B 1 is “solved” via scanning, editing and verifying [ABC+22]

[ABC+20]
B 1 is “solved” via scanning, editing and verifying [ABC+22]
A 2 is solved via capture with Controlled Natural Language
[ABC+20]

Italian Railways Workflow [ABC+
20, Figure 1]

20, Figure 1]
Hopefully our successors will not be building anything this complex
to understand our artefacts, but how sure can we be?

20, Figure 1]
B We can reasonably hope that our artefacts are fully formal.

20, Figure 1]
B We can reasonably hope that our artefacts are fully formal.
A But “strong legacy domain knowledge”?

Paris Métro Line 14 [Déh22]

1998 Opened; fully automatic; critical software verified by
Atelier B. major showcase.

This software (rather than the construction process)
was signed off, with no change process [Déh17,
Minute 14:30].

Minute 14:30].
2003,2007 Extensions opened (but already in the verified
software).

Minute 14:30].
software).
2010s Planning for new extensions, not in software.

Minute 14:30].
software).
Atelier B only runs on Sun workstations, ported to
Linux (but not necessarily 100% identical, and
previous software didn’t verify out-of-the-box).

Minute 14:30].
software).
2020 New extension, with new trains running 2× faster.

Minute 14:30].
software).
The new trains were 8 carriages, not 6, so not every
door opens at every station; significant modification.

Minute 14:30].
software).
The new trains were 8 carriages, not 6, so not every
door opens at every station; significant modification.
q
But it’s running formally verified GoA4 [Cue20].

Questions

Questions
1 Cobol: Do you keep programming in a sufficiently similar
language?

Questions
language?
2 Atelier B: Is your proof infrastructure still maintained?

Questions
language?
3 Ligne 14: Just because a theorem verifies automatically today
doesn’t mean it will verify automatically in 20 years time. Is it
worth/possible recording some major steps of the automatic
proof, rather than “relying on upwards compatibility”

Questions
language?
This may be more serious as more machine learning
(sledgehammer etc.) is used in proof systems.

Questions
language?
4 Italy: Are you reliant on “strong legacy domain knowledge”?

Questions
language?
4 Italy: Are you reliant on “strong legacy domain knowledge”?
“The Empire can’t maintain anything” — Hober Mallow
[Asi44].

Bibliography
I
E. Ábrahám, B. Becker, A. Bigatti, B. Buchberger, A. Cimatti,
J.H. Davenport, M. England, P. Fontaine, S. Forrest,
D. Kroening, W. Seiler, and T. Sturm.
SC2: Satisfiability Checking meets Symbolic Computation
(Project Paper).
In Proceedings CICM 2016, pages 28–43, 2016.
A. Amendola, A. Becchi, R. Cavada, A. Cimatti, A. Griggio,
G. Scaglione, A. Susi, A. Tacchella, and M. Tessi.
A model-based approach to the design, verification and
deployment of railway interlocking system.
In Proceedings International Symposium on Leveraging
Applications of Formal Methods ISOLA 2020, pages 240–254,
2020.

Bibliography
II
A. Amendola, A. Becchi, R. Cavada, A. Cimatti, A. Ferrando,
L. Pilati, G. Scaglione, A. Tacchella, and M. Zamboni.
NORMA: a tool for the analysis of Relay-based Railway
Interlocking Systems.
Tools and Algorithms for the Construction and Analysis of
Systems TACAS 2022, pages 125–142, 2022.
I. Asimov.
The Big and the Little [reprinted as “Merchant Princes” in
Foundation, Gnome Press, 1951].
Astounding, 33:7–54, 1944.
A. Cimatti.
Migrating Railway Systems.
Personal Communication at Dagstuhl February 2022, 2022.

Bibliography
III
O. Cuenca.
Paris inaugurates Line 14 extension.
https://www.railjournal.com/passenger/metros/
paris-inaugurates-line-14-extension/, 2020.
D. Déharbe.
Talk at SCSC Summer School.
https://vimeo.com/showcase/5271198, 2017.
D. Déharbe.
Évolution de la ligne 14.
Personal Communication 11 August, 2022.

Bibliography
IV
Financial Times.
Ulster Bank’s computer says no to negative rates.
https://www.ft.com/content/
ff12856a-e07d-11e9-9743-db5a370481bc, 2019.
E. Shein.
COBOL programmers are in demand to fight the coronavirus
pandemic.
https://www.techrepublic.com/article/
cobol-programmers-are-in-demand-to-fight-the-coronaviru
2020.

Bibliography
V
S. Sumner.
Computing gathered a group of UK-based IT leaders together
to discuss their legacy concerns, including security, skills and
cost issues.
https://www.computing.co.uk/news/4057289/
cios-discuss-legacy-issues-dont-write-mainframe,
2022.
M. Thomas.
Ethics, old software and negative interest rates.
Letter to Financial Times 4 October, 2019.

Have we a Human Ecosystem?

More Related Content

Similar to Have we a Human Ecosystem? (20)

More from AdaCore (20)

Recently uploaded (20)

Have we a Human Ecosystem?