Modern Relics

The National Institute of Standards and Technology hosted a workshop in March on
long-term knowledge retention, looking for answers to the questions of what digital data government, industry and academe should be saving, and how it should be saved.

They came up with no immediate answers, but they confirmed that a problem certainly exists and it's growing, said Josh Lubell, a computer scientist in NIST's Manufacturing Engineering Laboratory.

According to estimates offered at the workshop, the world churns out enough digital
data to fill the Library of Congress every 15 minutes. Much of that is of no interest to anybody and can be discarded quickly. But in areas such as engineering, the production of information is outpacing our ability to ensure
it will be available to those who may
need and want to share it later on. As
computer-aided evolves into computergenerated,
terabytes of data are disappearing,
or becoming inaccessible or corrupted
every day.

'So much information is digital, and
people are feeling the pain of losing access
to their information,' Lubell said.

The attendees at the workshop agreed
on the need to establish a business case
for long-term data archiving and to develop
standards for ensuring
interoperability
of data across time
as well as across
hardware and software
platforms.

Specifically, engineering
and design
drawings are routinely
saved today, said Doug Cheney, an interoperability
consultant and product director
for ITI TranscenData of Milford,
Ohio. But unlike blueprints of the past,
computer-aided design drawings, which
are becoming increasingly important elements
of geospatial initiatives, represent
only the tip of the design iceberg.

'Digital data is not hardened, especially
3-D geometry,' Cheney said. 'It is very interpretive,
so it's easy for unexpected
things to creep in.'

The problem is compounded, because
changes that creep in when data is moved
from one platform to another often are
not conspicuous.

'It's what we don't know that will kill
us,' Cheney said.

This dilemma came to light in the aftermath
of an industrial accident in Green
Bay, Wis., when a pipe feeding a boiler exploded,
scalding two men to death. The
immediate problem was corrected, but in
the next few years the company involved
changed hands several times, said Crispin
Hales, an engineering forensics investigator
in Winnetka, Ill. In the process, a lot of
institutional knowledge was lost.

'They almost had a repeat of the failure,
and they had no records of what had happened
only five years before,' Hales said.

Tacit knowledge goes missing

The problem is the disappearance of what
Hales called tacit knowledge. Forty years
ago, organizations were stable, people remained
on the job for years and there was
a community of knowledge that could be
tapped.

'That is gone,' Hales said. It has been
replaced by a more fluid environment,
where access and exchange of complex
data has replaced knowledge.

At the same time, data has become
more complex. Too often that data, if it
was retained, is not reliable because the
systems used to generate it
are no longer available.

'The government has this
problem in spades,' said
William Regli, associate professor
of computer science at
Drexel University.

Large engineering projects by
the Defense and Energy departments
generate huge
amounts of data on systems
that quickly become obsolete.

NASA's space shuttle program,
one of the most complex engineering
projects in history,
kicked off in 1981 and the last
shuttle is not scheduled to be
retired until 2009. The B-52
Stratofortress, for years the
backbone of the Air Force's nuclear
armada, first flew in 1954,
and its operational life span is
expected to extend to 2040.

The preservation and maintenance
of electronic data is a
problem many agencies are
wrestling with, including the
National Archives and Records
Administration and the
Library of Congress. The library has for
several years been digitally preserving information
from other media, from audio
recordings to parchment manuscripts.

Now, under the National Digital Information
Infrastructure and Preservation Program
authorized by Congress in December
2000, it's working with other federal agencies
and private-sector organizations to develop
a national strategy for collecting,
archiving and preserving the growing volume
of materials created only in digital formats
[GCN.com/584].

But NIST and others are particularly
concerned about preserving the engineering
data that goes into missions throughout
government. The Library of Congress
and NARA funded some of Regli's work at
Drexel University on digital preservation.
'I got into the engineering side of it
about five years ago,' he said. 'The digital
CAD process has a wealth of information
that isn't saved in any meaningful way.'

Engineering data had become too complex
for humans to process on their own,
he said. In the recent past, drawings produced
by a draftsman with pen and
paper were based on a human level of
knowledge. Computer-aided designs are
based on levels of complexity and assumptions
that are not reproduced when
the final drawings are printed or displayed.
These often ephemeral assumptions
and calculations can be as important
as the drawing.

Beyond our comprehension

CAD systems today deal with higher orders
of geometry beyond our comprehension,
Cheney said.

'You can't write an equation for the
geometry,' he said. 'The geometry is closely
tied to the software that interprets it.'
That software changes through a design's
lifecycle as it moves from one platform
to another.

'Each time it is produced, it's slightly
different,' he said.

One solution offered now by ITI TranscenData
is a CAD add-on tool that analyzes
potential shape and fit problems in
CAD models as they are translated between
systems.

Industry is aware of the problem of design
drift and corruption, but the solutions
so far are mostly industry- or even
company-specific.

'There is a lot of redundant effort in
American industry,' Regli said. This translates
into wasted effort and interoperability
problems. And it does not address the question
of long-term retention as systems become
obsolete. 'How do you ensure
interoperability across
time?'

These issues often get little
attention in the private sector
because they don't produce
revenue.

'The ultimate beneficiary is
not the current business unit,
but some future entity,' Regli
said.

Current shareholders often
are reluctant to pay for things
future shareholders will bene-
fit from, Lubell said.
That is why government is
getting involved in developing
technical standards for longterm
preservation. But government
will not be able to solve
the problems by itself.

'The business case has to be
made first,' to convince people
to use standards, Lubell said.
'It's a challenging area because
it's not dealing with an
immediate issue,' Regli said.

Standards today

There are some standards in
place, most notably the International
Standards Organization's
Standard for the Exchange
of Product Model
Data. But none answer all the
needs for interoperability
with long-term integrity and
availability.

Within government, NIST's Manufacturing
Engineering Lab has an exploratory
program for developing data standards,
and the IT Laboratory has a digital
preservation program that is investigating
how digital data degrades over time.

Efforts are under way by industry groups
to extend STEP for archiving quality by
adding geometry exchange specifications,
and to include definitions for validating
accuracy when data is exported. The European
aerospace industry has begun a standards
process for archiving 3-D images
that at least some of the U.S. aerospace industry
have joined.

Once adequate standards exist, the
question will remain, what needs to be
saved?

As storage becomes cheaper and access
quicker, Regli said one philosophy is:
'Why not just store everything?' But
years down the road, that approach could
produce the equivalent of digital archeology,
as engineers and researchers sift
through terabytes of data searching for
the nugget they need.

'You have to strike a balance,' Lubell
said. 'One of the big challenges of longterm
archiving is deciding what to keep for
how long. We're not doing a very good job
on that now. There's a long way to go.'