Monolithic structures

[jducoeur]

Thinking back to yesterday's posting about the current brouhaha over Windows Vista, and having read more of that thread, I'm quite struck by the consistent tone of the postings -- almost everything posted by engineers indicates that the problem is that Microsoft's internal development process has gotten so overwhelming that it is completely impeding all progress.

And y'know, I'm starting to wonder if the problem is the concept of Windows. Not whether it's well-implemented, not whether Microsoft is good or evil, not whether it is carrying too much legacy baggage, but simply the fact that it is, at its heart, designed to be a monolithic monster of a system. That's the whole point: Microsoft sustains itself by building more and more stuff into the OS, fending off competition by building the system bigger and bigger. And I suspect that that model is now crumbling.

It's fascinating to note that some parts of Microsoft are thriving. Reports indicate that the next version of Office is going to be quite nice, and I know that the programming-tools side is doing decently well for itself. Indeed, what's really striking is that some parts of Vista that were excised into separate, version-neutral projects look like they're now likely to ship before Vista -- and I'll bet they're more solid for it as well. When things are run as separate products, they're still capable of making progress, and even (occasionally) innovation. It's just the giant monster that is Windows that has ground almost to a halt.

I find myself drawing an analogy to the USSR. In the end, the main reason the Soviet Union fell wasn't because it was evil, or because it was imperial, or any of that: mostly, it fell because it didn't work. Central planning doesn't scale: it can work for a small organization, but the larger things get, the more inefficient things get. Microsoft seems to indicate that the same is true of software -- even the biggest, richest software company around just can't run a single project that's this big. Yes, there are many contributing causes, and everyone is pointing the fingers at them, but I suspect that they're really side-effects: I'd bet that the real root cause is that the entire idea is bogus.

Prediction: if Microsoft is still thriving in ten years, it will be because they have broken up the OS, at least internally if not into legally separate units. Right now, it's following the same path as AT&T and IBM, getting sclerotic as it gets bigger and bigger. And like those companies, I suspect it needs to be shaken to its core if it's really going to get healthy again. Ironically, they may have done themselves a disservice by fighting the anti-monopoly suits as hard as they have -- it could have served as just the shock the company needs...

Comments

[dsrtao]

RST and I were throwing around that idea yesterday -- that the only way to get a really big OS distribution is to do it in a decentralized fashion. In the open source world, there's an organization that does kernels and one that does the big C library and compiler, another that handles the windowing system and four different groups offering major mail-transport-applications -- none of them order the others around, and while they lose something in coordination and competition, they gain much more by not having to get permission, by not seeing stupid mandates from management, and in general, by focussing more on "what's right" than "what does marketing say we need?"

Coordination

[metageek.livejournal.com]

I think the open-source world gains from having lower coordination; it means that nobody has the right to tell the browser folks that they have to change to be compatible with the latest version of the kernel.

[goldsquare.livejournal.com]

I think you are close - but the real thing is that Microsoft has a piss-poor architecture, has interpenetrated one product into another, has hidden side effects, and lots of undocumented features that people have come to rely upon.

They, for competitive advantage, deliberately abandoned every known process for intelligent software engineering. It is a tribute to them that they got as far as they did, freighted with millstones as they are.

Now, what do they do?

[jducoeur]

An interesting question. For most companies, I'd simply predict doom, as the architecture continues to go down the drain. I'm not quite willing to do that in this case, because Gates *has* demonstrated the ability to turn the ship -- I'm thinking of both the moment when he realized that the Win95 architecture was fatally flawed, and when he finally "got" the Internet. In both cases, the company managed a fairly impressive (albeit slow) transition, due to the extremely strong mandate for change from the top.

*Can* it be done this time? I'd say it's possible if and only if Gates gets scared enough in time. This is an even harder and huger problem than the previous two, requiring either an unparalleled refactoring of the codebase or simply throwing the whole thing out and starting again. If he becomes convinced that it's necessary for corporate survival, and realizes this at least 3-4 years before it *is* necessary, then I can imagine succeeding.

*Will* that happen? I dunno. I don't see the urgency at this point: while their market share is eroding, it's doing so subtly, as PCs gradually become a less dominant lifeform in the computing world. Microsoft is doing decently in those fields, but is nothing like the dominant player it is in the PC environment. It's quite possible that Windows will simply become *irrelevant* over the next decade, as our concept of a computer system changes; whether Microsoft can survive in an environment where it has to actually compete remains to be seen.

It's an interesting topic for some scenario planning -- trying to figure out the factors that are going to play in, and seeing how they interact...

[goldsquare.livejournal.com]

They have so much money - they can and will do all of the above.

Vista will turn out to be the costliest software Turkey ever. And they'll remain profitable.

[(Anonymous)]

Microsoft should model themselves after General Electric - proof that an enormous company can thrive. General Electric is well-known for having the absolute best managers in the world. Once you are a manager there, you are required to attend GE managerial schooling throughout your employment. And they trim off the dead wood every single year - no matter what position. Improve, or get out. It sounds like Microsoft takes the exact opposite approach. Tenure rules. Bah.

The simple fact that Balmer has the position that he does right now is utterly astounding. That fact alone fortells a brackish doom.

[jducoeur]

From your mouth to God's ear -- my mother-on-law has essentially her entire assets in GE stock. (Yes, I've tried to get her to diversify. Haven't succeeded: her money all comes from that stock, which her husband accumulated over decades at the firm, so she prefers to leave it all there.)

That said, AFAIK even GE hasn't ever tried to run a *single* project that was as central to the company as Windows is to Microsoft. That seems the essence of the problem: I suspect it's just too big and too monolithic a project to be feasible, even if it was well-run. (Which it appears not to be.)