; -
4
Mi:

I miš

rogrammed.
Powerful new Often, the attempt to convert is killed
|, concepts and | by the people controlling information:
technigues are | ' ' systems (IS) finances. The conversion
| becoming avail-.' ., process itself creates no new applica-
able to migrate.«  ' tions. Management perceives a great
the billions of - deal of effort and expense with nothing
| lines of existing . to show for it. There is a long and seri-
" code into more. :.. ous application backlog.
bm easily expand- :'...' In one organization after another, iri-
A : able, maintain-/",, ' cluding some of the most prestigious
able systems.  '': data-processing organizations, the at-

'— In most enter-
prises, computing

4 tempt to make a major conversion has
failed. Yet, the cost of maintenance in

Mi ARTIN systems have the non-I-CASE world escalates daily.
žirketa large guantities ' One of America s best; telephone
of old code, most : switehes ran into so He software-

built by hand in an unstructured fash:

| ion. The data usually has not been mod:
eled, with little or no interest in achiev-
ing data pom pali between dif ferent |

Sko n BM

" These systems typically were designed
-. before today's principles of good data-

— v base design were understood, The data .'
isn't normalized and is unrelated to the
data-administration process. As caretak- .
ers of these systems know all too well, " :
these systems are ikluke Riki Ga PM hi

to maintain. H
Traditional maintenance of programs '
is an unsatisfactory and expensive pro- ''!
cess. It has been compared with the at: '':
tempt to repair a wooden boat at sea:': |
New planks can be replaced only by us-.:
ing existing planks for support. The pro-
cess must, be done in small steps or the:'"
boat will sink. And sooner or later, the -.

boat. must be brought to a URE zra

rebuilt. ': | |

Using computer- aided si P VATE engi-'

neering (CASE) tools, old systems can «,,.,
be rebuilt from restructured design in-.
formation stored in the CASE reposi- '.
tory. They can't all be rebuilt guickly, "
because this would involve too much |
work. The best that can be hoped for is
a steady, one-at-a-time migration of the:
old systems into the cleaniy, znplnectedii
jorm.

In corporations that have tuli:
implemented development methodolo-""?
gies based on integrated CASE (I-CASE) |
tools, these are two development worlds. '  oralirjii)s!
The I-GASE world has the ability to nah Minnufvš are
evolve systems, continually improving o sis de! i
their design and regenerating code. The ''.'
systems are cleanly engineered, easy to; '
change and have stable data models.

Alas, there is also an underworld of ge mle
poorly structured old systems lacking ';//'. !' ' cost of malntalnin
data models. in many corporatlone, RIT | prooa bli
more programmers s their time" či: iii? 5 33" REKI TTEES
maintaining poor-guality code than TINA. s | jasi
those who work in the I-CASE world. </.... maintenance difficulties that its mainte-

' The problem is rather like the slum-  ..— nance costs exceeded $1 million per day.
clearance problem in a city. Despite a že The switeh would never have been built
city's new center, elegant architecture ' if that figure had been forecast. '"

and efficient street plan, existing A recent study by the U.S. Air Force

Ho ovakine kia ME estimated that unless maintenance pro- :
A maintained. Planners hope fora k a Ku Soc viy is improved by the year 2000,
neki ngrajcn from the slums and .'.,.25 percent of the U.S, draft-age popula-

W eli-d | s [don will be regulred to mainjala who Atr |

. Force' s software! ||:
Many za ke sa VE arab a."  Fortunately, new tool sets to facili-
major conversion of a file system to a . tate the complex process of rebuilding
database system and have failed. Usual- systems are available, The goal is to re-
ly, the reason is that the project con- "

—" verse-engineer old systems into a
sumes much more work than anticipat« ". cleanly structured form, using tools
ed because so much has to be re- that automate the tedious parts of this

Old systems riddled
with undocu-
mented and —
unstruc-

tured

code

New

New

systems

based on

an integrated-
CASE life cycle

iu UVE
: KS: AMIS NEPRI

' i T
4 FI ! ah, 1 s
k 4 4 | -
kas mah di j
Mile xi
"«

Li - ed '
t s si k jHe<; h"
4 i H |

Bo m. kia: o. h H J pI p- ETA"
Na, ii SN na ae a ii a:
po A O a O a UVA

ga PPP OO PA a RATA Ma |

4 ! m | j ;
E s H % k si se:
5 ij Na uh ; i ti ] ge . ,
A AE aii PC WEEKNAPPLICATION
kon," - i ki oi i | | | h |

NEUUREČEV

DEVELOPMENT

APPLJED INTELLIGENCE

 Reyerse: Engineering Gives Old Systems New Lease on

process. The programs are captured
and represented in a CASE format so
they can then be modified as reguired
using CASE tools. Data definitions are
captured and converted into stable

data models stored in the CASE repos-

itor

The figure shows a reverse: engineer-
ing step. The code of the old system is
restructured with an automated tool
and entered into an I-CASE tool so it

can be analyzed and redesigned, and be-

come part of the I-CASE evolutionary.
life cycle.
In connection with this reverse-engi-
neering step, three terms are used:
Restructuring. conversion (with an

How To Reverse-Engineer Existing Systems

Old Systems Can Be Rebuilt Using Restructured Desigii
Špecifications Stored in the CASE Repository

viš

IE Old code is
testrieluted with an
dut6mated tool and entered into

an integrated-CASE tool where it is analyzed and redesigned.

ma zna one ee ponnanoik:

UŠI; s] Analyze ;
| /

|

code

| John Avakian

stadij A kodi ! BaA ala dxt—t

In one a ne after another, the attempt.
vr to convert to a database system has failed. Yet the
utal old systems escalates daily.

au tomated tool) of unstructured code

into fully structured code.

' Reverse-engineering: conversion of un-
| tucied code into high-level design
' specifications and (automatic) entry of
— these specifications into an I-CASE tool
€ where they can be improved or re-

| Če gaja

Re-engineering: modification of the de-
vsign of a system, adding functionality
where reguired, and (automated) pro-
duction of code for the improved '

system.

' If reverse-engineering is done as

A art godec pa je zene
rebuilt as evolutionary systems. Eventu-

Lite

ally, the messy underworld should dis-
appear.

There are basically three approaches
to the IS reverse-engineering and re-engi-
neering problem:

6 Do Not Convert. Allow applications
to continue their existence unconverted
but, when necessary, build a bridge to
new systems using data models and pro-
cess models stored in a CASE repository.
6 Restructure. Guickly restructure the
messy applications (preferably with
automated tools), but don't rebuild them
with CASE tools. The slum areas are im-
proved, but not rebuilt to be part of the
new planning. Often it's necessary to
build a bridge to the new systems built
— with CASE tools.
6 Rebuild. Reverse-engineer the old ap-
plications to conform to the data models
and process models stored in the CASE
repository. This is comparable to rede-
signing and rebuilding areas of the city.
In large IS installations, a mix of these
approaches will probably be used.

To Convert or Not

Two guestions should be considered
before deciding to convert an old sys-
tem. First, does it work? If it works
well, there's a strong argument for leav-
ing it alone. If it works inadeguately, it .
should be rebuilt using integrated CASE
technology. Second, does it incur high
maintenance costs? If the application is
fragile and expensive to maintain, then
it's a candidate for restructuring or re-
building with automated tools.

It may be appropriate to automatical-
ly restructure an old system and modify
it by programming rather than rebuild
it with reverse-engineering and I-CASE
tools. If an application system works ad-
eguately and needs little maintenance,
then its conversion should probably be
postponed. Spend the effort: on some-
thing else; there are so many other ap-
plications needed, such as a bridge be-
tween it and the new environment.

In planning data resources, it may be
unwise to assume that old systems will
be converted easily to the new, automat:
ed form. A realistic appraisal is needed
of the costs and difficulties of conver-
sion. The dismal history of uncompleted
conversions should be weighed.

In their initial enthusiasm for new de-
velopment methodologies, system design-
ers often assume that the old systems
will be converted and discover too late
that they won't. It's safer to assume
that many old systems will survive and
plan a bridge that links them to the new
world.

Next week | will examine the chal-

' lenge facing vendors in building tools

that num reverse-engineering. —

once] S embodied in reverseengi:
neering are described'in the CASE vol
ume in The James Martin Report Se-
rtes. For more information on this vol-
ume, call (800) 242-1240. For
information on seminars, contact (in
the United States and Canada) Tech-
nology Transfer Institute, 731 10th St,,
Santa Monica, Calif. 90402 (213) 394-
8305. In Europe, contact Savant, 2
New St., Carnforth, Lancs., LA5 9BX
United Kingdom (0524) 734 505.

SS RTU mni) aa naje pa aa m di
Win s. — niaš a ič
zao Mla - Hi