|
INDUSTRY WATCH: Two Sides to Every Transformation
By David Rubinstein
 January 15, 2004 — Phil Murphy of Giga Information Group was reading the Oct. 15 issue of SD Times when he came across my story about Object Management Group’s effort to create a specification for transforming legacy applications and systems to bring them up to date.
“I thought, ‘I’ve gotta rebut this somehow,’” Murphy recalled of the article (“Breathing New Life Into Legacy Systems). “It’s like communism. In theory it sounds good, but there’s always a ruling class somewhere.”
In a report he released on Nov. 3, Murphy took issue with the article, outlining the reasons he believes this legacy transformation effort is, in his words, fatally flawed. Among them are that OMG has bitten off too big a piece to swallow, that legacy transformation isn’t cost effective and the gains are relatively minor, and that the entire effort is in his view not pragmatic at all. He said he would recommend to his clients that they not go down OMG’s path.
The OMG effort, according to task force co-chairman Bill Ulrich, involves more than moving data off mainframe systems into more modern architectures. Indeed, it encompasses “any production-enabled system regardless of language or platform.”
Of this all-inclusive approach, Murphy said, “They’re going after the world. They don’t do any of this in one language, but now they want to take on them all.”
Murphy believes some programming languages should be left to die, languishing in a Darwinian struggle as unfit to survive. “Vendors make tools that go after languages on the upswing, [because] that’ll make them money,” he said. COBOL is still alive, he said, because it is closely aligned to business functions. But by embracing old languages like Fortran, PL/1, LISP and others, Murphy claims, OMG’s project is unworkably broad in scope and doomed to fail.
A bigger problem, according to Murphy, is that the legacy transformation process can’t be automated. Therefore, he believes, a company could get a quarter of the way through one of these huge transformation projects, only to suddenly realize that all the data is out-of-date. “Unless there is a dynamic collection vehicle, they’re just building a mountain of data no one will trust,” he said.
But the biggest nail in the coffin, according to Murphy, is that the “magical mythical description layer” OMG proposes be built over every piece of an organization’s technology stops at the paragraph component level. “Then, you have to manually document what [the code] is supposed to do. But what if I don’t understand it enough to document, or if it changes, the static documents’ comments are out of date.”
When asked about Murphy’s comments, Ulrich said the group hasn’t made a final decision yet as to whether the next RFP will capture code-implemented processes. “We didn’t want to tackle that in the first RFP,” he explained.
The goal of the legacy transformation project, Ulrich said, is to give a higher view of a system, to help organizations sort through and eliminate fragmenting, redundancies and inconsistencies that create inefficiencies in their systems.
Ulrich said the automated gathering of data from different systems in different languages already has been done. He cited a project he ran for the Internal Revenue Service in the 1990s, allowing for the information to be coalesced into a common view.
As for the language issue, Ulrich said, “The fact [Murphy] believes Fortran is useless is his opinion. There are 13 million lines of Fortran code, mission-critical for financial services institutions. The whole Darwinian issue—you can’t say one-off systems must die. PL/1 is small compared to COBOL, but it’s big around the world. It’s not our position to judge” which shall live and which shall die.
ANOTHER APPROACH?
Murphy said he’s a proponent of application portfolio management, using existent impact analysis tools, as a better way to solve the problem. The market, he said, has proved that the “sausage grinder” approach—COBOL in, Java out—has failed. “Service orientation is the answer to the CORBA-Java problem,” he claimed. “Wrapping screens always has been a ‘hold-your-nose’ deal, but those 3270 screens are closely aligned to business functions.”
Ulrich, though, claimed that simply wrapping screens and creating Web services in different parts of an organization won’t help it gain any efficiency. “Then, you’re just propagating the problem.” He said that assuming systems don’t need to change and that organizations only need impact analysis tools “is wrong. If wrappering doesn’t fulfill the need, or you can’t rewrite [a piece of code], then this could be an alternative.”
Ulrich did acknowledge that his OMG task force is changing its name due to an apparent misconception as to what legacy transformation means. “We’re not restricting it to COBOL,” he said.
The new name that the group decided is less ambiguous is the Architecture-Driven Modernization group. You be the judge.
David Rubinstein is executive editor of SD Times. |
|
Advertisement
Click here for a complete listing of Industry Watch Columns
Click here to see a complete Column Archive. |
Back to Top
|