One of the odder habits of the native Minnesotan is becoming cold voluntarily. As evidence, I offer the St. Paul Winter Carnival.

Not wanting to deprive my children of a True Minnesota Experience, we trundled off to this remarkable event. The parking was free, but a mile away — this is a popular event! — so we trudged through the snow (against the wind and uphill, of course) and availed ourselves of the festivities: We admired snow sculpture (in warmer climes, you’d use sand, I imagine), traversed the “ice maze” (you haven’t lived until you’ve wandered through walls of ice blocks containing seaweed and an occasional dead perch), and most importantly, obtained mini-donuts, kettle corn and hot chocolate.

At the end, I asked the kids if they were cold. “It has to be absolute zero out there, Dad!” exclaimed Kimberly, the elder of the two. Never passing up an educational opportunity, I asked if she knew the temperature of absolute zero in Fahrenheit. “Sorry, I can’t convert from Celsius to Fahrenheit in my head.”

Okay, then what’s absolute zero in Celsius? “Stinkin’ cold!”

Kimberly isn’t the only one who appears to be unsure how to convert concepts to numbers. Following my article last year critiquing Gartner’s “The Cost of Migrating COBOL Developers to Java,” I received a polite note from one of the authors, J. Feiman, who, seeking to refine my understanding of his opinion, sent me another half-dozen Gartner reports on related topics as evidence of the completeness of their analysis.

I read every page, but found nothing to change my assessment. While Mr. Feiman assured me he’s in favor of retraining COBOL programmers, he provided nothing to refute any part of my critique, which pointed out that while every imaginable cost of retraining COBOL programmers had been quantified, of the cost of recruiting Java programmers only salary and benefits had been turned into numbers. Mr. Feiman didn’t respond to my invitation to call and discuss the matter. InfoWorld has made an open offer to print a Peer-to-Peer column, should Gartner choose to submit one; to date that offer is also still open but unaccepted.

In his final message Mr. Feiman expressed distress at my disinterest in his opinion. It’s a fine distinction — I’m more interested in what he’s published than what he thinks, because that’s what IS managers will use when making decisions.

Then there’s the letter I received from Vincent Maiello, Practice Manager for the Professional Services Division of Chubb Computer Services:

“Our experience has been more in line with your comments. Chubb Computer Services is a premier training and staffing organization and subsidiary of The Chubb Group of Insurance Companies. We have found that major organizations find re-skilling their employees to be more cost effective than hiring new employees, not only because of the cost of recruiting and ramp up time, but because of the (non-monetary) value of the employee’s knowledge.

This “human capital” approach has been very successful with some of our biggest clients. In fact, today we are running seven concurrent classes for one of the world’s largest life insurance companies re-skilling their mainframe folks to web developers. When we piloted the program earlier this year, the client’s response was great and the employee morale and retention has increased.”

Chubb Computer Services is in this business, so it may have an axe to grind. On the other hand, it’s also in the staffing business, so I imagine it wins either way.

(Caveat: I don’t have the ability to research and compare competing offerings vendors. That’s why I generally avoid any mention of products or services — doing so might appear to constitute endorsement. Mr. Maiello’s comments, based on first-hand experience, were sufficiently relevant to this discussion that I made an exception, although I have no idea whether Chubb’s re-skilling programs are better, worse, or equally effective than any others.)

If you’re thinking about re-skilling your COBOL workforce, here’s what you now know: Gartner thinks it’s a good idea, despite its analysis purporting to show that re-skilling is the more expensive choice. My unrefuted analysis of their analysis suggests their numbers aren’t worth the magnetic domains they’re stored in.

But that’s theory vs theory. The experience of a re-skilling practitioner — in this case demonstrating successful conversion of COBOL jockeys to OOP practitioners — outweighs any theoretical analysis.

Especially when it agrees with me.