In Larry's Kingdom (Part 2)
The second half of the 90s: Internet Explosion!
(This is a continuation of an earlier story)
In Part 1, I had led an Internet Engineering Task Force group to produce RFC 1697. This was a win, and I also led the group to create the Oracle-specific extension to that. We helped create an “app” for HP Open View, which was also way fun. Some people from the HP group at Oracle (at the time, HP was the leading platform for the Oracle database, so they were a very important group) and I went to Atlanta to demo our app at an HP show. I got invited to speak at system management conferences. Life was good.
This was good for me personally, but it wasn’t Big Money for Larry. What Larry wanted was something to sell to Oracle customers, and that was the product line ultimately called Oracle Enterprise Manager. This was software you’d install somewhere on your network, and use it to manage all your databases . It wouldn’t need to manage or even monitor an Informix or an IBM database like OpenView could — just Oracle’s.
Unlike SNMP, on which I didn’t really have much serious competition within Oracle, Enterprise Manager was a political football right from the start. Lots of people wanted to do it. This is a good place to mention “capitalism vs. socialism,” which is probably not what you were expecting in an article about Oracle. Nonetheless:
The US is a capitalist economy, for the most part. Corporations compete with each other. Actively cooperating to divide up markets or fix prices has been illegal since the late 19th century (that doesn’t mean no one does it, of course).
But inside the corporation, it’s usually quite different. Tasks are allocated, charters given, and competition inside the company is strongly discouraged. “Good management” means strict division of labor, and “clear lines of responsibility” are universally praised. In fact, this is so taken for granted that I couldn’t even find a business school publication that says it (I’m sure if I looked harder, I could find some).
It’s instructive to see what this economics website says about state communism:
Under state communism (e.g. Marxist-Leninist) , the Communist State gains control over the means of production and decides – what to produce? How to produce? and for whom? Examples include the Soviet Union and Eastern European states. Under state communism, there is a high degree of centralisation with production targets set by central committees and local officials being responsible for meeting these targets. States with a Centrally planned economy have frequently involved considerable degrees of political control. In the 1930s and 40s, the Soviet Union had periods of rapid industrial expansion and economic growth. However, the system increasingly became bureaucratic, wasteful and lacking in incentives. Towards the end of the Soviet Union, the economic system was increasingly broken with major shortages of key goods and service
“Under state communism, there is a high degree of centralisation with production targets set by central committees and local officials being responsible for meeting these targets.” — does that sound like the Executive Committee’s budget meeting at your company? A “centrally planned economy” is exactly what most corporations aim for.
Let’s suppose, under Communism, you are one of those “local officials” and you discover that another local official is poaching on your territory. What do you do? Assuming he doesn’t have an official sponsor higher up in the Party, you probably ask your own Party official to make him stop it or else.
That’s exactly what happens in most corporations. Your division has the charter to produce System Management, and no one else is allowed to undertake it. There are clear lines of responsibility and no “wasteful duplication.”
You’ve probably figured out by now that that isn’t what happened at Oracle. Oracle was internally capitalist, more than any other place I’ve worked. It was perfectly fine to go into competition with another division, up to a point (that point where top management finally decides who wins). It was not considered a faux pas like in most companies
In this case, another division (the Desktop division) created its own “management agent,” which was the software that ran on the machine with the database, communicating with the central manager. It was called BattleStar. It used Tcl scripts to do most of the work, which was actually a pretty clever idea. I appropriated that.
I hadn’t even heard of Python then, which, with benefit of hindsight, would have been an even better choice. Perl would also have been good. But Tcl was having its moment in the mid-90s.
Oh noes! We thought System Management was our job. The battle was joined, no pun intended. I’ve since heard that Larry actually thought that having two management platforms competing with each other in the market might be a good thing! However, he ran into some disagreement with his other top managers, so that view didn’t prevail.
Most of the war took place at pay grades way above mine, so the bare-knuckle stuff was handled for me, thank God. Anyway, we won, and I got to manage the Enterprise Manager agent, which included the SNMP agent and a lot of Oracle-specific stuff. The BattleStar team were assigned to “help” us, and let’s just say I treated them as defeated enemies, and leave it at that.
Managing in the Dot-Com Boom
This was in 1995 or 1996, and the dot-com boom was well underway. It seemed like everyone was getting rich, and if you weren’t, you wanted to start.
Hiring people became next to impossible, and unfortunately a very large part of being a manager is getting people to join your team. Jay Smith, the guy I mentioned in Part One of this, decided to move to New York to be with his girlfriend, and I spent about three days with him getting his Sun machine to work without a network, and getting a leased line to his apartment. Eric Bosco, a fantastic guy whom I had hired out of the Oracle Desktop division, left and went on to (much) greater things, starting at AOL.
I transferred a woman from Oracle Dublin, who was earning (I’m not kidding about this), $35,000. We raised her to $50,000, earning us a note from Larry asking what percentage increase that was (at the time, Larry had to approve all Engineering hires).
I hired a guy from another company, and we paid a lawyer $10,000 to have his H1-B visa transferred to Oracle. Then when he told his employer he was leaving, they convinced him to change his mind and stay. I told them they should pay Oracle the $10,000 he cost us, and amazingly enough, they sent me a check in the mail! Engineers don’t normally handle money, so this was a novel experience for me.
As if all that wasn’t enough, Oracle management imposed a new source control system on us, ClearCase. It is now owned by IBM, which should tell you something: IBM only buys products that have to be sold to vice presidents and above. No one anywhere near the trenches would ever think of using ClearCase.
So we had ClearCase, plus a layer of software that Oracle had put on top of it. ClearCase was a Unix system that was twenty years ahead of its time: you installed software on your Unix machine that created a virtual file system representing the “correct” versions of every file that was under source control. What could possibly go wrong?
A lightning lesson about dealing with computers: when everything goes to hell, a file on a hard disk is usually the one thing you can count on. Now, even that was gone. Life was hell.
At my next company, there were two of us from Oracle, and we threatened to quit if they brought in ClearCase.
On top of the hiring and the ClearCase debacle, my Enterprise Manager agent was a first release, and I’m sorry to admit that it was not very robust. Or so I discovered when it went to the platform groups.
At Oracle, we wrote “portable” C code, according to a very rigorous coding guide, and then it passed over to the platform groups, who ported it to Sequent or DEC or HP or SCO Unix or whatever. Anyone in Oracle who was having trouble getting the agent to run would dial my number. It seemed like they all were having trouble.
Managing the Agent sucked. In fact, managing itself sucked. When I had just two terrific guys working for me, it was great. Now? Not so much.
I transferred to a group that was doing a Web server. Was this a cowardly act? Absolutely. I’m not going to whitewash it. The going was getting tough, and I got going.
The Nerd Case
In 1997, a major lawsuit erupted between Informix and Oracle. It was tried in a Portland courtroom, and Portland, at the time, was very unused to tech companies suing each other. If you asked the staff in the courthouse what room the trial was in, they said, “Oh, you mean The Nerd Case?”
Eleven Informix developers in Portland, led by their vice president Gary Kelley, left Informix en masse and joined Oracle. They were apparently very unhappy with Informix’s corporate direction after it bought another database company, Illustra. Gary and crew had been shopping themselves to other companies, including Microsoft.
Informix had been making a lot of noise about how they were way ahead of Oracle, technologically. They had a billboard alongside 101 right by the exit for Oracle, telling drivers they had just passed Oracle, and so did Informix. Yawn. “We let them pretend they’re competing with us” was the Oracle attitude. Databases are a strategic, long-term decision for a corporation, and they don’t make it based on one or two features.
The story had some juicy bits for the tabloids, such as Informix president Phil White walking over to Larry’s house in Atherton late at night, personally, to confront him about it (Larry was out of the country at the time).
Informix sued, claiming theft of trade secrets. Oracle claimed that it had no interest in their “trade secrets” and I have every reason to think that’s true. I had plenty of personal knowledge of the “Kernel Group”, that group of developers who built the guts of the database, and they were as much Royalty as any group at Oracle. I’m absolutely certain that they had their own plans for adding features similar to Informix’s and had no interest whatsoever in talking to Gary about it.
At any rate, it came out in the pre-trial hearings that this was true, and Oracle was not stealing any trade secrets. Oracle agreed that Kelley’s group would not work on any technology based on their Informix experience, and Informix dropped the suit.
So What Will They Do?
Now Oracle had this new group of developers in Portland, where it had no engineering presence. This was a decade before the Portlandia show that spoofed the city’s weird culture, but I think we can assume the things it satirized were present even then.
These people thought extremely highly of themselves but couldn’t work on that one thing that was their main claim to fame.
Now an Aside: Weasels
Weasels are members of the Mustelid family, which includes carnivorous animals like badgers, ferrets, otters, wolverines, and many animals that we tend to like (except for farmers and people who keep chickens). Yet “weasel” is the one animal in that group that’s become an all-purpose insult. Urban Dictionary defines it as a “ shifty, schemeing [sic] person that will do whatever they need to to escape whatever they fear in the moment.” Weasels are the villains in many Walt Disney films.
Personal aside: in Engineering, we call it “going over to the Dark Side” when an honest, hard-working engineer moves into marketing or top management. Whereas in engineering, you’re (ideally) rewarded for being right and accomplishing things, on the Dark Side you’re rewarded for sounding right and promising to accomplish things. This is why engineers hate marketing people and VPs: they can’t actually do anything, but they certainly can talk about doing things.
Top executives are often regarded as weasels by ordinary engineers. This is an AI-generated photo of one:
And Now Back to Oracle
I had known Gary Kelley around 1985 when he was just an engineer at 3Com. He’d apparently gone over to the Dark Side with both feet. He set to work in the top management circles of Oracle, selling his group as God’s gift to Silicon Valley. Informix apparently had this whiz-bang database technology that he pioneered, but he was legally enjoined from using it. What was he going to do?
He tried to take over Oracle Enterprise Manager, but my old manager was way too crafty for him. I believe weasels, who are actually not bad creatures but do have that reputation, would be offended to be compared to Gary.
And so, they became our bosses in the “application server” division (or whatever it was called). Somehow, Gary had weaseled himself into this. I had several trips to Portland as a result. My immediate manager was someone from Portland.
The upshot was, we hated them and they hated us. My buddy Mark had a “competitive lab” where we stocked machines from our competitors so we could try them. Gary shut it down.
I began looking at startups myself. The Web was not nearly as well populated in 1997 as it is now, but there were still a fair number of websites. One good thing about having a manager who’s not around, at least back then, was that you could spend your time on whatever you wanted without fear of your manager suddenly dropping in.
One of the sites I looked at was http://pw.com (which apparently doesn’t exist anymore). The “pw” was for “Price, Waterhouse” and it was a database of startups. I looked at hundreds of them, and they all seemed to be some variation on “We’re using the web to disrupt <old industry>” or “We’re providing hardware and software for all those companies that are disrupting old industries.” Sometimes it was way more nebulous than that. It was often not even clear how they’d actually make money.
One of the startups, called “Packeteer,” had a refreshingly unpretentious pitch. They were providing hardware and software to optimize your bandwidth over TCP. I read the founders’ biographies, and they were technical people without Ivy League MBAs. I thought, “This sounds like a reasonable way to make money.” I didn’t know a single soul there, but I sent an email, one thing led to another, and I left this New Yorker cartoon on my office wall: