IT Archaeology: A Word from the Dig Site

Impact LESA BULLETIN 2002-005

IFPTE LOCAL 28 AFL-CIO & CLC

by Dr. William H. Jones

Administrator O’Keefe has related on several occasions that when he first announced to his family that he was to be named Administrator of NASA, his child remarked “Gee, Dad, I thought you had to be smart to work at NASA.” I would not want to be accused of being overly solicitous of the Administrator, but I think we might all agree that he can be safely credited with having raised a relatively sagacious child; while being smart may not be, technically, a requirement for working at NASA, it is at least a very good idea.

This time around it is not Administrator O’Keefe that needs a gray-matter infusion, but instead his newly-appointed Chief Information Officer, Paul Strassmann. As I am a computer engineer of some 35 years experience, things that Mr. Strassmann says ring a little louder for me and, so far, nothing he has said has particularly impressed. Having read a number of Mr. Strassmann’s remarks, let me suggest that one of the first requirements for a Chief Information Officer ought to be that he know something about how to program a computer. It would be nice if he had written a computer program somewhat beyond “Hello, World!” within the last computing millennium or so. (A computing millennium lasts 6 to 10 months, depending upon the current rate of technological advancement.) I have been over Mr. Strassmann’s personal web site and find no record that he has ever programmed a computer, in any millennium. He writes a lot, and consults, and advises; now as NASA’s CIO, he may start dictating, but I find nothing about him ever having actually pushed the keys down himself. A doctor may be quite eminent, but if he has never actually cracked a chest before, I would prefer that he not start with mine.

Apparently, Mr. Strassmann recently took a tour of the Information Technology aspects of NASA and described it as like “an archaeological expedition”; not an entirely flattering remark, but perhaps he said it to a crowd of employees summoned to a parking lot somewhere, which would bring it up to NASA’s standards. One thing Mr. Strassmann might want to educate himself on is the small budget issues that have been eating NASA alive for the last few years: it is something of a challenge to revamp a center’s computing strategy while the space station is moaning “Feed Me!” Also, he might interrupt his chanting of “One NASA, Good NASA” long enough to notice that NASA is, in fact, quite diverse, with quite diverse computing needs.

We range from the people at the Cape who fill the tanks with fuel and push the big red button to the folks in Cleveland who are trying to figure out how to get the blue light to come out of the warp nacelles. The same dumb terminal fiber-coupled to his new computing center at Marshall is not going to meet all those needs. And finally, I have to wonder just who Mr. Strassmann visited on his archaeological expedition. I can state for a fact that he didn’t visit me: if he can look at what I am doing and see it as fossilized footprints in the creek bed of computing, then one of us has our plug out of the wall, and it is not me.

While we have passed by dumb terminals fiber-coupled to a computing center at Marshall, it is internally reported that that is in fact his plan for IFMP implementation: dumb terminals (cleverly called “thin clients”) with card swipers connected by dedicated fiber from hither and yon to a glass room at Marshall; talk about architecture that was hot in the 70s and died in the 80s, where has this man been? In an age when everyone else under the sun is moving to distributed computing, the network is the computer, intelligent agents roaming about, object-oriented yada yada yada, this man is buying System 360 from Big Blue and enclosing it in plexiglass! This is the sort of thing that leaves you grasping the arm rest of your chair desperately feeling for the ejection handle.

Another thing Mr. Strassmann is pushing forward is his idea of Mission Control Centers for the NASA network. There may be some few things that are appropriately centralized, but these must be considered with extreme caution. As the patriarchs learned back in the early 70s, centralizing most things net is a bad, bad, very bad idea. Indeed, the founding design goal of the net was to make it decentralized and redundant so that it would continue to provide reliable communications literally through the progress of a full-scale nuclear war. Mr. Strassmann wants the Mission Control Centers to watch every computer, every router, every bridge, every port, even every keystroke, and let them know when any one, any where on NASA’s net so much as twitches. I have news for him: there is a whole lot of twitching going on and his fools-designate are going to drown in it. And if there is a twitch, Mr. Strassmann thinks Mission Control is going to get that nasty person off in a microsecond. This brings us to two more truths he doesn’t seem to know. In a microsecond, light can travel only 982 feet; that might, as the fiber flies, get you out of the Mission Control building and across some of its lawn. Cleveland is still some 8,000 microseconds away. And even if he uses that fancy blue light they’re developing up in Cleveland and gets there in that microsecond, at three billion instructions per second (this week, more next week) a lot can happen.

Another lesson that anyone who has been programming for most of the morning knows, but seems to have escaped Mr. Strassmann, is that even small, insignificant changes in the computing business can have disastrous consequences; indeed, it is those small ones that are usually the worst. Mr. Strassmann seems to have some distaste for static IP (Internet Protocol) addresses, I suppose because they require someone to keep a data base of them. I hear he wants to move to dynamic identification through DHCP (Dynamic Host Configuration Protocol). DHCP has been around, it works; that shouldn’t bother anybody. Well, it is going to send my work for a loop or two: my work builds upon CORBA (Common Object Request Broker Architecture) and CORBA uses static IP addresses from here to there. Why? Because they are fast and they don’t change. Key feature. If you would like to talk to that object again a month from now, you just slap that static IP address down on the counter and give it a call. If the machine is still there but has changed its address because DHCP gave it a new one last week when it rebooted, well you might as well pull its plug because its objects are lost to you and your silly old static IP address.

Mr. Strassmann tells us that IFMP (the Integrated Financial Management Program) is going to get NASA’s budget, and in particular Space Station’s budget, under control because everybody will be counting beans into the same set of piles. Ha! I will admit that all 10 field centers and headquarters count beans in different ways and into different piles and that that isn’t a terribly convenient thing for people who want to look over the big picture, but that isn’t why Space Station is skidding out of control. As the august committee report said, one of the key budget difficulties with Space Station is that they still can’t tell you what a complete Space Station has: how many nuts, how many bolts, how many modules, how many computers, whatever. Furthermore, they don’t really know how long it is going to take to build however many things they decide make a space station. They manage to their yearly budget: they spend all they get and then decide that that is how much they were supposed to do for that year. And another thing they don’t know is how many years the Russians are going to expect Space Station to pay for the Russian space agency. Just a few weeks ago the Russians suggested that they were out of money. They said they weren’t going to make or fly any more Soyuz spacecraft and they wanted the one they had up there back. End of Space Station. I fail to grasp how consistent bean counting is going to bring that under control.

And while we are on IFMP, Mr. Strassmann might want to take the training before he trumpets it too much more. As a computer engineer of long experience, I have to say that IFMP is embarrassing. For such a Rube Goldberg hodgepodge of COTS (Commercial Off The Shelf) software to be presented as NASA’s flagship example of government accounting makes me want to hide in shame and tell people that I work for a secret adult entertainment complex in Los Angeles. That there is a place that you should never click because we don’t use that feature and all you’ll do is hang a task in the system or lose your work is bad enough, but that there are two or five or ten such places is appalling. That we are inventing new paper forms to cover up for the fact that the COTS software doesn’t do things that our old systems took care of is even worse. That this is perhaps the best that the likes of Mr. Strassmann and the headquarters crowd can do may be true, but if this is the best that NASA can do, we might want to take Russia up on her offer to bring those last three folks down from the Space Station after they turn out the lights.

Now what do I want? Do I want Mr. Strassmann strung up by his neck on the front steps of headquarters? Well, a public hanging now and then is fun (as long as you are not the guest of honor) and does generally lift the spirits of the troops, but perhaps it has become politically incorrect in this day and age. And I suspect that Mr. Strassmann is a far more competent and reasonable person than his reported comments make him appear, just as Mr. O’Keefe appears to be a reasonable and competent person who is decidedly against the parking lot approach to management pronouncements. Maybe Mr. Strassmann assumes that everybody knows he means just business systems, that technology and mission computing is in a different kettle entirely; unfortunately, those qualifiers aren’t getting reported. Perhaps as a compromise Mr. Strassmann might just promise to study some advice that John Kennedy often quoted: don’t tear down a fence until you know why it was put up.

Dr. Jones is available through e-mail at William.H.Jones@grc.nasa.gov and also reads the grc.talk newsgroup regularly; however, he reserves the right to say nothing at his convenience.