In the spring of 1995, I had an opportunity to witness a historic occasion: the removal of a firm’s mainframe IBM computer. I wrote a couple of columns for Infoworld back then and was recently reminded about the moment. Here is what I wrote:

The notion of turning off your mainframe computers is somewhere between romantic and impossible, depending upon whom you talk to you. Our esteemed editor-in-chief gave an actual date (this month) when the last mainframe at Infoworld would be shut off — that is the romance part — while I am convinced that there will always be room for mainframes in corporate America, somewhere.

The company that I visited is Edison Electric Institute, the trade association for the nation’s several hundred investor-owned electric utilities. They are located in downtown Washington DC and had — until recently — a single IBM mainframe computer running their business.

At the core of any trade association are lists: “We are constantly putting on conferences, mailing things, and updating records of people in our member’s companies,” said Jon Arnold, the IS director at Edison. Edison publishes over 200 different handbooks, conducts training seminars, and even tests potential operators of nuclear plants.

Edison had originally purchased a 4331 about 12 years ago. “At the time, we were spending hundreds of thousands of dollars with an outsourcing company — only back then we called them remote job entry sites,” said Jeff Kirstein, one of Arnold’s managers. “We had written all sorts of applications in COBOL and everything was charged back to the end user.”

Edison decided to buy their own mainframe and began to write what would end up being a core of six different applications, most of them written in SAS: a list management routine that had over 70,000 names on over a thousand different lists, a meeting registration package to keep track of hotels, courses, and other details needed for the various conferences the trade association put on each year, a financial tracking package, a package to keep track of the several hundred publications and periodicals that Edison subscribed to, and a committee membership tracking system.

SAS was flexible, to be sure. “But it was also a hog, and as a statistical package it was poor at generating business reports, but once we got the hang of it, it worked fine for our purposes,” said Kirstein.

The mainframe-based applications “got the job done,” said Jim Coughlin, a programmer analyst at Edison. “We had setup our mainframe to make it easier for each user to maintain their own lists — every mainframe logon ID was tied to a particular virtual machine and memory space. However, the mainframe systems were fairly crude: each time a user opened a list it would take time to resort and their was minimal error trapping during data entry.” One plus for the mainframe was that “we didn’t have to spend a lot of time running the machine,” said Arnold. Using IBM’s VM operating system, Arnold had one system operator and an assistant to do routine maintenance and backups.

But the annual fixed costs were high, and when Edison moved four years ago into a new building in downtown DC they began to deploy LAN-based technologies. At the time of the move they began thinking about turning off their mainframe. However, they had to manage the transition slowly:

“First off, my predecessor signed a five-year lease with IBM shortly before I came on board,” said Arnold. That lease, along with the monthly software maintenance charges for the various applications and system software, ran to about a quarter of a million dollars a year. Breaking the lease early would have been costly, so Edison decided to start planning on turning the mainframe off when the lease expired in February, 1995.

Secondly, they needed to find another tool besides SAS that would run their applications on the PC. More on what they picked next week.

Third, they had to populate their desktops with PCs rather than terminals. They began to do this when they moved into their new building, first buying IBM PS/2 model 50s. Now Edison buys Dell Pentiums exclusively. All of the desktops are networked together with two NetWare 3.12 servers, and run a standard suite of office applications including Word Perfect GroupWise and word processing, Lotus 1-2-3 and Windows.

Fourth, they had to retrain their end-users. This was perhaps the most difficult part of the process, and is still on-going.

I’ve seen lots of training efforts, but my hat is off to Edison’s IS crew: they have the right mix of pluck, enthusiasm, and end-user motivations to make things work.
The IS staff raffles off a used model 50 for $50 at a series of informal seminars (“You have to be present to win,” said Arnold). And “we also kept a list of people who didn’t get trained, and they don’t get access to the new applications on the LAN,” said Jeanny Shu Lu, another IS manager at Edison. Finally, the applications developers in IS held a series of weekly brown-bag lunches with their end-users, and solicited suggestions for changes. These would be implemented if there was a group consensus.

I turned off my first mainframe at 1:11 pm on March 2nd, 1995 in the shadow of the building where the original Declaration of Independence is kept. Somewhat fitting, but somewhat sad. I was at the offices of the Edison Electric Institute, who happen to be located across the street from the National Archives in downtown DC.

When I first spoke to Jon Arnold (the IS manager at Edison) last year about doing this, I was somewhat psyched: after all, I had been using mainframes for about 15 years and had never been in the position of actually being able to flip the big red switch (a little bigger than the ones on the back of your PCs, but not by much) off before. However, when the time came to do the deed on March 2nd, my feelings had changed somewhat: more of a mixture of bathos and regret.

IBM mainframes were perhaps responsible for my first job in trade computer publishing in 1986: back then I was deeply involved in DISOSS (now called OfficeVision and almost forgotten), 3270 gateways, and products from IBM such as Enhanced Connectivity Facility and IND$FILE file transfer.

The mainframe I was ending was a small one: an IBM 4381 model 13. It had, at the end of its lifetime, a whopping 16 megabytes of memory and 7.5 gigabytes of disk storage — DASD to you old-line IBMers. Since this may be the last time you see this acronym, it stands for Direct Access Storage Device. It was running VM, which stands for Virtual Machine, as its operating system. And since it cost over a quarter of a million dollars annually to run, it had reached the end of its cost-effective life. It was replaced by two NetWare file servers, one of which had more RAM and disk than the 4381 had. The network was token ring, reflecting the IBM heritage of Edison and the utility industry it represents. Edison took existing SAS and COBOL applications running on VM and rewrote them in Magic and Btrieve on the LAN.
Magic is an Israeli-based software development company that provides tools to build screens and assemble database applications that can run on a variety of back-end database servers. One of the reasons Edison began using it was they had a contractor who began to build applications using it. “However, they never finished the work and we had to take it over,” said Coughlin. “It is a wonderful rapid application development language, and easy to make changes.”

The combination of the two products is very solid, according to Edison’s developers and end-users. “We’ve had no data corruption issues, no performance problems at all,” said Kirstein. “We had tried other products, such as writing X-Base applications in Clipper, but that was a mess.”

The Magic-based applications took less time than SAS. For example, rewriting the meeting registration package took two months, about half the time the equivalent SAS application required.

And the new LAN-based applications allowed Edison to improve service to their members, avoid duplicate mailings, and increase the quality of their databases. “We forced more cooperation among our end-users since now one person owns the name while another owns the address,” said Kirstein. “We’ve added lots of user-requested capabilities to our list management software, but we’ve also managed to keep it centralized,” said Arnold.

Getting rid of a mainframe, even a small one like this 4381, isn’t simple of course. It took Edison several years to migrate towards this LAN-centric computing environment and to rewrite their applications on PC platforms. And then they still had to pay several thousand dollars to have someone cart the 4381 off their premises. (Their original lease, written five years ago, stipulated that Edison would be responsible for shipping charges.)

So in the end, the machine had a negative salvage value to Edison. They managed to make a few hundred bucks, though, on a pair of 3174/3274 controllers that someone wanted to purchase. Ironic, though, that the communications hardware (which was even older technology than the 4381 itself) would prove to be worth more and outlive the actual beast itself.

Turning it off was relatively easy: I typed in “SHUTDOWN” at the systems console (forgetting for a moment where the “ENTER” key is on the 3270-style keyboard), waited a few minutes, and then hit the off switch on several components, including a tape drive and CPU unit. Ironically, the cabinet of this mainframe was red: perhaps Edison knew they were going to replace the mainframe with a NetWare network long ago? The hardest part was remembering the command to turn off forever their 3270 gateway, which was a piece of software so old from Novell that it no longer was supported or manufactured (why upgrade something that you eventually will retire?).

What made the moment of shutdown a sad one was that I was standing amidst people who had seen the 4381 first come in the door at Edison, and who had spent a large portion of their careers in the care and feeding of the machine.
“It is going to be pretty quiet in our computer room,” said Kirstein.

After we shutdown the mainframe, the Tricord running NetWare was still running. And it was alot quieter.

Where are they now? Arnold is now Managing Director WW Utilities Industry for Microsoft.

Most of the time, you hear founders of startups talking about their great idea that is sweeping in scale and scope. But last night I went to a meetup where the exact opposite was proposed: an idea that will take you exactly 10 hours to put together: or, as the presenter Dylan Hassinger tells it, “one long night of coding.”

It bears further thought. Because as you flesh out your idea, you find out that it is going to take you a lot longer to really pull everything together. As an example, consider last week’s subject, Juristat. It took them the better part of a weekend to formulate the kernel of the company, and now, many months’ later, they are just seeing an actual product.

But the 10 hour idea is a good goal to keep in mind. The current term is “minimum value proposition” but this really strips things to the barest, most essential part of your startup. What is it that you are trying to get across? If you can figure something out in 10 hours, you have your elevator pitch, your have the essence of what you are trying to do. You have your hook for hiring new help. It gives you a lot of leverage.

At last night’s meetup, Hassinger boiled his own presentation on the “bulletproof startup” into a sound bite: “”Use your skills to build a tiny product with real value for a growing market that excites you.” That I like. He told us that he originally started out by saying an idea should take three months, then one month, then a weekend. I like that he arrived at 10 hours. One all-nighter. That has some appeal (although lately I value my sleep higher and higher).

Too often I see startups that are all over the place. They are trying to do too much too soon. They have multiple great ideas, and flip from one to the next, trying to satisfy the whims of their last mentor or last customer or the defects that they uncovered at their last coding session. Sure, there are a lot of blind alleys that you have go down as part of the startup process: that is to be expected. But if you have this laser focus, you can strip away everything else that isn’t essential for your actual business.

Sometimes, the 10 hour idea also means that you drop features from your product, because they are taking you away from what is that essential essence. It is easy to get into the feature war: just one more feature, really, I promise. I see this all the time in the software products that I review for major publications. They add and add and add features until you can’t find your way out of a menu tree. In review sessions where I face down developers and marketing folks (often the first time that they are actually in the same room together, which is sad and should be the subject for another column), they see how complex they have made their product and where they went astray. “But our biggest customer asked for that feature!” Sure thing. Then you add another. Resist the temptation to add complexity.

You can read my review in Network World here.

You can download the various screenshots here.

And you can follow the Twitter handles of the various vendors here.

A little over a year ago, Drew Winship was one of the participants in the St. Louis Startup Weekend. The goal was to find others of like mind and form a company within 50 hours over a weekend. I watched as he brought his team together and began to formulate their plan, and it was fascinating. Winship’s idea was to download information from the state court websites to determine the best days and trial judges for lawyers to try their cases in front of. Because there are differences among them, and if you know going into a trial that you can get a better outcome some place else, why not use that to your advantage?

For example, why not figure out the probably of a judge granting a summary judgment (meaning a shortened trial)? Or which litigators at particular firms have better won/lost case stats? Or if your stats are better than competing law firms’? Or the composition of particular juries? If you get enough data, the analyses can be pretty compelling, particularly for large dollar law suits.

There was just one tiny problem: the Missouri state court website didn’t have any easy way for the general public to download its data. So Winship and his team put together a series of automated scraping techniques to gather the data over that first weekend. That created something of an issue, because this automated scraping effort looked to the court website as a denial of service attack in progress.

So began Juristat. You could see how lawyers and their regulatory bodies might not have much of a sense of humor, especially when it comes to having their data taken without their permission. Indeed, if you examine the sad tale of Aaron Swartz and his suicide earlier this year, he was essentially doing something similar with downloading batches of academic journal articles.

Winship is a member of the Missouri bar and somewhat reluctant to become a cowboy data wrangler: mainly because he could be put in jail for these activities. The courts decide what is and isn’t appropriate fair use of their data, even though it should be available to the public. This reminds me of the early days of the Internet when folks like Carl Malamud fought with the US Patent Office and the Securities and Exchange Commission to to free their data archives. Now we don’t think much of having this kind of access: indeed, you would be hard pressed to find folks that prefer the paper documents to what you can find online.

I spent some time with Winship last week and he brought me up to date on his fledgling company. While Missouri’s courts have since balked about giving him any data, he has been able to legally access the entire New York state court system and process it in his system. He is expecting contracts from a couple of major Manhattan law firms “any day now” and has continued to develop and refine his algorithms to make them unique and useful to his legal clients. Part of their challenge was to develop application interfaces that would work and that others could use to build on top of Juristat’s efforts. And part was just manipulating the data in such a way that it would be useful for your average lawyer with no computer knowledge. (Insert your favorite lawyer joke here.)

Yes, there are some pretty big players in their space, including Lexis and Westlaw. They don’t have all the data that Juristat has, and they don’t have any accessible analytics either. That is the golden opportunity available, and why the company has won the attention of Arch Grants.

But what I like about Juristat’s story is that you don’t often find its founders that simulate a DDOS attack to start up their company. And certainly not by a bunch of lawyers! I wish them well.

The Internet has made it easier to stay in touch with people from our past: just this week I met with someone that attended my undergraduate college (whom I never met when I was there) and got an email from a co-worker from my past. It put me in a mood.

This job change was a big moment for me: it turned me towards my path of tech journalism and changed the nature of what I do every day.

I can remember it precisely: it was the winter of 1986, and I was working as an IT analyst for a large insurance company in downtown Los Angeles.

It was a fun job. Back then, end user computing was on the rise. Budgets and staff were big. We had, I think, somewhere north of 20 people working in various capacities, and we were installing PCs by the truck load across our three building “campus” (although no one called it that back then). I was good at my job, and enjoyed working with end users and helping them to learn about the few apps (by today’s standards) that we supported on their PCs.

I was an avid reader of PC Week, which (along with Infoworld) was the leading trade pub for IT workers. All work would stop when the internal mail delivered our copies Monday afternoon, as we tried to scan its pages before our users (who also were subscribers) would start calling us with questions about the tech they were reading about in the latest issue.

PC Week was starting a special edition that was going to be called Connectivity. It would be a supplement that would go to a subset of its readership, what publishers call a “demographic.” And they were looking for writers and stories.

I found out who was going to be running the publication and sent him what I now know is a query letter. At the time, I was just an interested reader of the pub and didn’t think anyone would be even interested in hiring me, let alone want to know what I thought was important and interesting. I mean, I was just this little cog in a big machine. I had zero professional writing experience. I didn’t know who the CEOs were of the major tech vendors by name. I was in the process of installing my company’s first LAN, so was interested in PC communications.

I was dead wrong. They were more than interested.

That query letter led to flying me out to Comdex in Vegas, the biggest tech show back then, and meeting the newly minted Connectivity staff, and eventually a job offer to join their ranks as a staff writer. I began working for them almost immediately, and the rest, as they say, is history.

That first year I wrote more than 300 individual stories for the publication. They were stories about how LANs were connecting to mainframes, and how PCs were changing the nature of American business. They were heady times: we had the ear of every major tech company around. I got to work with some of the most creative and interesting people of my career, some of whom I still am in touch with today. Many of the original PC Weekers went on to bigger and better things in the tech industry, and I am proud to count myself as part of them.

I went from installing my first LAN to telling thousands of people how to do it themselves. Back then, we could talk to anyone, including Bill Gates, just by calling them on the phone. Email was still a bright and shiny object, and the Internet was yet to be invented, by Al Gore or Vint Cerf or anyone else.

What motivated me to write that letter? I really don’t know. But it was a transformative moment for my career, to say the least. And I tell this story to you today in the hopes that you may share your own moment when your career went in a new and exciting direction, and you have the perspective to acknowledge and celebrate it. Please share your stories here if you feel so moved.

You can read my post on Slashdot here.

You can read my article here.

I have a long history of testing these tools. Almost seven years ago, one of the Techtarget publications had me looking at them for larger enterprises, and I went out to the central IT department at Stanford University to put them through their paces. This time around, I wanted to find a small business site for the tests that I was going to be doing for Network World. That’s why I was over at Mercury this past week.

They have about 10 Macs connected to an Apple Airport, which is the center of their network, providing IP addresses, wireless connections and a shared hard drive to the entire office. The Airport is attached to a cable modem and the Charter broadband network.

Wait a minute. Don’t you need a firewall if you are going to connect your network to the badass Internet? Yes, and Mercury knew they were taking chances. A firewall is just the basic separation that keeps the bad guys from getting inside your network and causing havoc. That is why they were the perfect testing site. They were vested in my review and what I would find out about these products and their specific needs.

Interestingly, it isn’t just small businesses that don’t have firewalls. When I arrived at Stanford, the central network didn’t have any either. Partly that was because of some odd notion of academic freedom, but back then they realized they had to get better protection. Ironically, while I was doing my tests there we saw someone try to reach out from Germany one morning. Luckily, they had other defenses that prevented them from doing any damage, but it emphasized the reason why I was there testing these products. And coincidentally, when we brought up the advanced firewalls at Mercury, we could see all the network traffic where folks were continually scanning and looking for ways to enter their network too. It was a sobering illustration of why these products are essential.

When I first arrived on scene, I went into their phone closet where I tried to suppress a gasp. Yep, this was your typical small business: part storage room, part cable jungle, and mostly a mess. It was clear that trying to figure out the network topology was going to be a challenge, and my first act was to leave everything alone.

Inside the closet were two small gigabit switches from DLink that looked like they had been around since the days of DOS. This worried me, but since things were working, I wasn’t too concerned. Yet.

One of the vendors that were part of the test insisted on sending a product engineer to help with my testing, and I am sure glad that he was there. When we cut over to his device instead of the Airport, things initially went south. Turns out we found a bug in their firmware. Once that was fixed, all of the wireless Macs were quickly brought up on the network behind the new firewall. But the wired Macs had trouble connecting. It took a few reboots later before we got everyone back on board. It was ironic that the wireless portion of their network was easier to bring up than their wired portion. That was thanks to the wonky cabling in the closet.

So what are some takeaways from this experience?

If you are running gigabit Ethernet to your desktops, make sure your cable plant is up to snuff. Part of my problems had to do with the older cables used to connect things in their wiring closet. There is a difference between Cat5 and Cat5e, especially if you want to run the faster networks these days. Make sure you are using the right cables.

Disconnect any unused wired ports in your office.  This is just basic security practice, but bears repeating. And if your wiring contractor hasn’t done so, you should label your ports in the walls and in your closet so you can track things down more easily.

Understand the limitations of your core network gear, including switches, routers, firewalls, and wireless access devices. Your network installer should explain these things in terms that you can understand.

Have a separate guest network with the appropriate security measures. The Mercury folks were using the Airport guest network features, which were bare bones. One of the reasons they wanted to go to the advanced firewall was to provide better protection from their frequent guests and contractors who were going to be connecting in their offices.

Oh, and what happened with my review for Network World? Well, you will have to wait and read about it in their pages. I can tell you that I learned some interesting things about all the products that I tested.

I spent some time talking to the folks from NASAtv who broadcast up to six different channels of programming during the event. These guys are amazing in how much they know about the various technologies to pull off broadcasting in harsh conditions. You might not think that having a TV truck in the middle of what usually is a football stadium is hard, but consider that they have to set up their own control room on what is usually the 35 yard line, and with a network topology shown above that has to be assembled in a matter of a few days. All of the competitions make use of wireless technology that is carefully tuned to ensure that each bot has a clear channel to its controllers and can only play on a particular field.

I took some pictures of the participants and their bots. One thing that I didn’t really grasp until this week was the wide range of equipment that the teams bring with them: the smaller kids in the Lego league mostly had fun decorating their pit stop areas, while the bigger bots (that can weigh more than 100 lbs.) bring in an entire machine shop as they have major surgery and last minute repairs to perform. Of course, there were fun costumes, hairdos, and other wild add-ons for the human participants.

Here is a Pinterest gallery of my photos. And here you can find a video that I shot two years ago.