Mind Your Language

28 APPLICATION DEVELOPMENT ADVISOR q www.appdevadvisor.com
D
EBATES on programming languages
are always sure to inspire more heat
than light. A previous article
looked at a number of languages
quantitatively as opposed to qualitatively,
concluding that once a language has made it
into the mainstream, it tends to be quite dif-
ficult to kill-off quickly or completely, and
that the industry is generally multilingual.1
As an industry it would be fair to describe
software development as being strongly
fashion-led. As with fashion, there is the per-
ception that whatever is at the shows and in
the magazines is what everyone is doing. As
with fashion this is not the case: people
still wear their old clothes and get on with
their lives. The call to “wake up and smell
the JavaBeans”2
because Java is the answer,
everyone is doing it, and C++ and all other
languages are in free fall is about as far
from reality as the view that Java is an
Internet language for a minority of pro-
grammers. From the other corner of the
debate, the fact that a language may be
better (whatever that means) than another
language has never stood in the way of it being
unsuccessful in the market place. The ele-
gant Eiffel is a classic example of this; the mar-
ket is a harsh antidote for idealism.The truth
is always a subtler blend of opposite views,
and tends to have less regard for fashion, logic
or conservatism.
Silent majority
It is easy to fall into a polarised view based
on one’s work environment.There are many
COBOL shops where the idea of GUIs
and curly bracket languages seem distant,
faddish, and irrelevant; there are vibrant ecom-
merce companies steeped in XML, Java
and Perl to the degree that there is almost
denial there has ever been another purpose
for software; there are Microsoft-only sites
where the idea that other viable operating
systems exist, and are measurably more
reliable, is at best hearsay; there are places
in academia where there is still a strong
belief that formal methods are viable con-
tenders for the mainstream; and so on.
The message is that software development
is more multicultural, in this respect, than
many individuals or consultancy groups
would like to admit—it is difficult to take
a position of extreme moderation and be
heard.
For many programming languages the con-
cept of the silent majority applies.There are
many languages that are definitely not fash-
ionable but are still quite widely used, away
from the glare of publicity. They are used in
the context of continued development of exist-
ing systems—euphemistically called main-
tenance by many, an unfortunate and
inappropriate term that software borrowed
from the physical engineering disciplines—
and, perhaps surprisingly, for the develop-
ment of new systems. With the year 2000
boundary condition out of the way, it is worth
taking stock of some of the better known.
Fortran
Fortran started life to prove a point: a
compiled language could produce machine
code with efficiency close to that of hand-
crafted assembler. John Backus led his
team at IBM in this endeavour between
1954 and 1957. He freely admits that
they made the syntax up as they went
along, preferring to focus on a close match
between the language and the features of
the underlying IBM 704 on which it was
originally developed. This certainly explains
something of the language’s syntax and
its frequent lack of consistency.
Although it has long since outgrown its
original acronym (FORmualTRANslation
because, at the time, its syntax had a vague
resemblance to mathematical notation) its
efficiency and ability to be first to market
for the numerate disciplines ensured that it
would win out of the more elegant ALGOL
58 and ALGOL 60.
Fortran has evolved radically through a
number of standards and extensions. Modern
Fortran is radically different to the FOR-
TRAN of the early 1960s (and not simply
because Fortran 90 sanctioned the lower case
spelling of the name). Programming using
exclusively modern Fortran features leads to
Enterprise Development Solutions
Now that the year 2000 crisis has subsided,
Kevlin Henney finds how legacy languages will fare
in the future, and charts their past development
MindYour
Language

differences comparable to the difference
between Java and classic C.
It was once said in the scientific pro-
gramming community that whatever language
they were using in the year 2000 it would
be called Fortran. It would appear that this
prophecy has been fulfilled. Fortran is still
taught on science and engineering courses
around the world, and is used extensively with-
in these communities. Although Fortran has
enjoyed popularity in commercial pro-
gramming domains, it has been largely dis-
placed leading to the mistaken perception
by many developers in the industry that it
has completely disappeared. The greatest
threats to Fortran’s position in its tradi-
tional market have come from C, C++ and
tools such as Mathematica.
COBOL
When people talk of legacy systems, COBOL
is without a doubt the language that most
often springs to mind. Along with Fortran
it is the language that one can most often
expect to find in the history of programmers
over the age of 30. It is not simply a mat-
ter of history, though. Y2K raised many peo-
ple’s awareness of how much COBOL was
out there. Although distinctly unfashionable,
COBOL code globally constitutes the most
code in any single language. Estimates vary,
but in terms of lines of code we are talking
international telephone numbers.
What we now know as COBOL started
life as an interim report (the final one was
never completed) resulting from a user
push for a business-oriented language under
the auspices of the US Department of
Defense and the newly formed Conference
on Data Systems Languages (CODASYL)
committee. Because of the perceived urgency,
the interim report was published as a stop-
gap measure in 1960 until the ‘real’ COBOL
could be completed. The final report never
happened.
Because of the vacuum for suitable busi-
ness languages, COBOL rapidly took hold
of this niche in the 1960s. Once there, it
proved difficult to displace regardless of
critique of its elegance, design, and other
attributes—“sticks and stones...”. Only
with significant changes in development plat-
form and software architecture has COBOL’s
dominance been challenged—PCs, client serv-
er, GUIs, object-orientation and theWeb have
all threatened it. However, in these cir-
cumstances COBOL has, through propri-
etary extensions, been extended to cope.
Although these have not led to anywhere near
the same market presence as the COBOL
mainframe association, they are still sig-
nificant. Through such adaptive extensions
and a splintering of mainframe dialects,
there are many COBOLs out there, and not
simply the ones defined through ANSI
standards.
What is COBOL’s future as the Y2K
teams wind down? Although some com-
panies have chosen to redevelop their sys-
tems to a new architecture and to a different
business model, there is less incentive to
throw it out of the many environments
where errant code has been fixed. The
language is still taught on many computing
courses, and has refused to be displaced
by more transient 4GLs, which have them-
selves caused upgrade problems. While the
demand for COBOL may drop a little, it is
unlikely to do so drastically.
PL/1
PL/1 (Programming Language 1) was orig-
inally touted as the last programming lan-
guage. As is often the case with such
ambitions, it certainly wasn’t, finding itself
in the company of thousands of other lan-
guages during the 1970s inflationary peri-
od of language growth.
It was conceived in the mid-1960s as a com-
plete general purpose programming lan-
guage, equally suited to scientific, business,
and systems programming. As such it used
the ALGOL, FORTRAN, and COBOL of
its times as starting points. A further criti-
cism often levelled at it was that because it
was many languages rather than one, it all
too easily allowed developers to program in
their original language’s idioms. It has been
said that one can program FORTRAN in
any language, and PL/1 certainly enabled this
to happen in a more complete sense than
most.
PL/1 was developed by IBM. Originally
named NPL (New Programming Language)
it was further developed at the company’s
Hursley labs in the UK, where it was felt that
the name might cause confusion with the
National Physical Laboratory.
SIMULA killer
The only other language at the time with sim-
ilar ambitions was Algol 68, which became
lost in complexity and committees. Usable
implementations of Algol 68 only began
appearing in the mid-1970s, long after the
original definition. It was competing with
PL/1, whose implementations were improv-
ing after a shaky start in the 1960s.The back-
ing of IBM and an ANSI/ECMA standard
in 1976 saw to it that in spite of its signif-
icant complexity, PL/1 stayed afloat while
Algol 68 persisted only in memory and
language design influence. PL/1’s com-
plexity led to the formation of subsets,
including PL/C and the official 1981 ANSI
standard PL/1 subset.
PL/1 was always poised for the big time,
to the point that it was used as the second
language in the classic Elements of
Programming Style (the other language in that
being FORTRAN 66).3
Its touted posi-
tion as the only programming language
one would ever need was responsible, in
Norway, for preventing major public fund-
ing of SIMULA, the first object-oriented lan-
guage. However, with the possible exception
of ALGOL, PL/1 did not displace those lan-
guages that it was intended to supersede. It
has been used typically for large systems, and
never fully broke away from living on main-
frames and minis. IBM sites were the most
likely to use it, and so its original status as
a vendor-created language was also never fully
shaken off. Its use has certainly declined in
favour of either newer or other old and
established languages, such as FORTRAN
and COBOL.
C
C is the progenitor of the curly bracket
family, which includes C++, Java, scripting
languages such as Perl, and interface definition
languages such as OMG IDL. It is descend-
ed from ALGOL 60, via the BCPL rather
than the Pascal lineage. Based on his expe-
rience of the Multics project, KenThompson
at AT&T experimented with a number of
languages, including BCPL, for developing
the kernel of what was to become Unix.
Dennis Ritchie developed C in 1972 as a
high-level systems programming language
for this purpose, rather than for the gener-
al purpose role in which it has since found
favour.
The fate of C and Unix were intertwined
in the early 1980s, but the uptake of C as
a general systems programming language
spread it far beyond the Unix environ-
ment, displacing assembler and other high-
level languages as the standard for systems
programming on many systems.The 1990s
saw C become accepted as general purpose
programming language, and it became a typ-
ical sight on developer CVs, either as the lan-
guage of choice or as a common second
language. C is a small language, with only
a handful of keywords and concepts.
However, some of these concepts and syn-
tax rules are tricky for the newcomer (and
even the experienced developer, as wit-
nessed by the International Obfuscated C
Coding Contest). This has often made it a
difficult language to move to for develop-
ers experienced in languages based on a
fundamentally different mindset, such as
COBOL.
The original de facto standard, the “White
Book”, came out in 1978.4
This is commonly
known as K&R C or Classic C. ANSI
JANUARY–FEBRUARY 2000 29

became involved in standardising the lan-
guage, leading to the C89 standard, and ISO
become involved soon after. A new standard,
C99, is about to hit the streets.The empha-
sis on portability has served to create a very
narrow dialect range for the language.
Compared to other languages, initial stan-
dardisation had little to do in the way of
accommodating radical differences between
compilers, and focused more on precise
exposition. Recent standardisation has
focused on extension, mostly in the areas of
internationalisation and numeric pro-
gramming. The former is a natural conse-
quence of the expansion of software markets
and the last follows from a sense of com-
petition with Fortran.
Classy C
The most successful divergence from regular
C is what was originally known as C with
Classes, and later renamed C++. This began
in the early 1980s and features in the lan-
guage influenced the standardisation of C.
C++ is largely compatible with C, which
helped it gain a foothold in the areas tra-
ditionally dominated by C; where C has led,
C++ has inevitably followed. For instance,
C has made inroads into the embedded
systems community, which traditionally
shunned higher-level languages, and C++ has
followed, leading to subset specifications such
as Embedded C++.5
Although C++, now an
ISO standard language, extended C in the
directions of safety, object-orientation, and
generic programming, it still retains its C her-
itage. It represents a challenging move to OO
for many non-C programmers.
However, C has not been completely
displaced by C++ or its more differentiat-
ed offspring, Java. Partly because of the mass
of existing code already written in C, and
partly because it holds a place as the low-
est common denominator most powerful
language available.The Open Source move-
ment of Linux, Perl, Apache, and others, and
the original free software movement, exem-
plified by GNU, is largely C-based. QA
Training’s experience of C is that the
demand for the language is still high,
although the profile of course delegates
has evolved during the last decade. Many
Visual Basic developers wish to get to grips
with C to better understand and interact with
the Windows API, and C++ developers
wish to find out more about the legacy
inherited from C.
In the The End of History and the Last
Programming Language, Richard Gabriel
forwards a theory for what makes a successful
programming language.6
C fulfils all of the
criteria; in one form or another it seems to
be here to stay as a living language.
Ada
Ada was a language born of committees, the
US Department of Defense, and Pascal. It
was a product of the 1970s and the endur-
ing belief that a final programming language
could still be created. Following a recog-
nition of the rising cost of embedded soft-
ware development, the DOD commissioned
a Higher Order Language Working Group
that was to identify requirements for DOD
languages and use them to evaluate the cur-
rent set of languages, which was large,
given the pre-Cambrian language explosion
of the 1970s. The result of the work was
that no existing language was entirely suit-
able, and work began on developing a
new one based on ALGOL 60, Algol 68,
Pascal and PL/1. Ada—after Augusta Ada
Byron, Lord Byron’s daughter and Charles
Babbage’s assistant, and allegedly the
world’s first programmer—became a stan-
dard in 1983, and a new standard was
approved in 1995.
Coming from the Pascal tradition, the syn-
tax of Ada is simple and readable. However,
Ada’s design tried to please all of the people
all of the time. The power of the features
crammed into its syntax often cause it to be
characterised as a complex language. It is very
easy to use a simple subset and be unaware
of the more advanced features, but in most
programming contexts simple features alone
are not enough.
Unfamilar model
Because Ada was created as a general pur-
pose language without having the benefit
of evolution to temper its features, some
of the concepts introduced in the first
round, Ada 83, were awkward to use or
insufficient. Ada 95 saw a host of new con-
cepts and features added to the language
to meet these perceived shortcomings,
some of which represented different lan-
guage models. For instance, although
object-based in its original incarnation, it
lacked many of the facilities, such as
inheritance or any form of dynamic pro-
cedure dispatch, which would qualify it as
object-oriented. Ada 95 introduced a
more complete object model, significantly
making Ada the first internationally stan-
dardised language to include OO fea-
tures. However, the model introduced
did not build on the existing task model
for concurrency, the newer protected type
features for monitor-like types, or the
common form for objects used in other sta-
tically typed languages. Although power-
ful, this makes Ada 95’s object model an
unfamiliar one to the majority of OO
developers.
Ada’s basis in defence work has ensured
that it still has a home there, but it is by no
means the exclusive choice for all such
software. The overhead associated with
many of its features and its runtime support
system has held it back from smaller embed-
ded systems use. Its reputation for safety has
led to its continued use in larger embedded
and safety-critical systems. Ironically, it
was Ada code that was used on Ariane 5’s
ill-fated maiden flight.7
Although this
expensive firework ultimately occurred
because of a failure to carry out appropri-
ate tests, there is a further irony that had the
system been written in C, the disaster
would probably never have happened! The
other significant market for Ada has been
in education, where it replaced Pascal in a
number of universities. However, the time
for even that market appears to be passing
with many universities favouring Java as a
first language.
Conclusion
Although not an exhaustive survey of old lan-
guages, this article has not been about find-
ing out whether or not a language is the right
fit for you, but how languages evolve and fill
niches, and how many programmers nav-
igate their own path between the Scylla
and Charybdis of fashion and pragmatism.
Certainly, when niches disappear we can
expect languages to follow, but although there
are always parts of industry undergoing
seismic shifts, it seems that there are also parts
of it that aren’t. s
Kevlin Henney is a Principal Technologist for QA
Training. He can be contacted at
khenney@qatraining.com
References
1. Henney, K. “Speaking in Tongues”,
Application Development Advisor,
2(1), Sept./Oct. 1998.
2. Guttman, M. Letters response to
“Java is an OO Language?”,
Application Development Advisor,
3(1), Sept./Oct. 1999.
3. Kernighan and Plauger. The Elements
of Programming Style, 2nd edition,
McGraw-Hill, 1979.
4. Kernighan and Ritchie. The C
Programming Language, Prentice
Hall, 1978.
5. The Embedded C++ home page,
www.caravan.net/ec2plus
6. Gabriel, Richard P. Patterns in
Software: Tales from the Software
Community, Oxford University Press,
1996.
7. The Full Report for Ariane 501,
www.esrin.esa.it/htdocs/tidc/Press/
Press96/ariane5rep.html
30 APPLICATION DEVELOPMENT ADVISOR q www.appdevadvisor.com

Mind Your Language

More Related Content

What's hot (16)

Viewers also liked (7)

Similar to Mind Your Language (20)

More from Kevlin Henney (20)

Recently uploaded (20)

Mind Your Language