- by Paul Murphy, -
I've been trying to imagine what the computing universe will look like in late 2008; an exercise which really drove home the degree to which Linux has become the focal point for anti-Microsoft feeling and underlined the risk that the Linux community now faces from commercialisation.
In some countries resentment against Microsoft's pricing, corporate arrogance, and shoddy products has become so co-mingled with anti-Americanism that pricing and functionality are no longer considered important for decision making. Recently, for example, price cuts and customisations offered in in countries as diverse as Thailand and Germany have almost certainly gone well past the level at which their acceptance would have exposed Microsoft to anti-dumping penalties, but had no effect on the customer's decision to go with Linux.
My belief is, therefore, that Unix, whether as Linux, BSD, or Solaris, will dominate the desktop everywhere except in the United States by the time 2008 rolls around. The really surprising conclusion, however, is that what will prevent Linux taking over in the US is expertise, not the relative price insensitivity of the American market or the absence of anti-Americanism as a sales driver.
Specifically, what I expect to see is that the expertise of the MCSEs now responsible for most corporate desktop computing will continue - meaning that they'll apply what they know from the MS environment to using Linux in both server and desktop roles. Mostly their efforts will fail to produce cost savings or productivity improvements because they'll be using Linux as a substitute for Windows without changing their own behavior or thinking about computing.
Trying to use Linux as a kind of second rate Windows is like trying to use house cats to pull a dog sled: you get a lot of exercise and frustration, but few positive results. Unfortunately, this has been the story of commercial Unix use for thirty years: a period of growing acceptance is followed by the widespread application of completely inapplicable expertise and consequent implementation disasters, after which the perpetrators quite rightly claim that Unix failed to meet their expectations and scurry back to the technology they know, badmouthing Unix all the way.
Current attempts to perpetuate the client-server architecture with Lintel desktops illustrate this perfectly. In business reality, the typical heavily locked down business desktop mainly runs applications which individually turn the PC into an expensive, failure prone, terminal to server based applications and file systems.
Replacing Wintel with Lintel achieves little; getting rid of client-server achieves a lot. Use central servers with smart displays like the new Sunray 1g and the user gets more power and control while the entire Wintel service delivery complex - from help desk to proxy server - along with all of its weaknesses, failures, and costs gets wiped out.
That has enormous positive impacts: Failures stop happening, security becomes a non-issue, all of the help desk and related support staff go away, and user productivity improves markedly. Application support focuses on the application, not the platform, and is delivered by peers who know the job, not bored help-desk people who tell the user to reboot.
That's technical reality, but business reality is that we're not seeing a large scale shift away from client-server. Ask the decision makers why not, and you'll find it's because the people making the decisions simply don't question the logic of using a fifty million line OS on a 3Ghz, 256MB desktop PC with 120GB of disk space to run a client consisting of an unreliable terminal emulator making insecure calls to local libraries. To them, that's the one and only natural way while anyone suggesting use of a secure and reliable smart display with a simple 10,000 line OS is both immoral and dispicable.
Push them a bit, and they'll generally insist that independent desktop processing protects continuity when the servers go down, but point out to them that servers shouldn't go down, that desktop users can't read email or run Word on documents from unreachable servers, or that the overwhelming majority of Wintel failures take place on the desktop and you may get baffled hostility, but you won't get change.
Does their position make make sense? No. Is it expensive for their employers? Extremely. Will it continue? almost certainly. Why? fundamentally because they believe they're doing the right thing.
They know that they're doing the right thing because their ideas about deploying technologies were learnt through experience and are tried and tested in the Wintel environment. Things like desktop lockdown, server stored files, storage area networks, dhcp, gateways and proxy servers, are the right things in a Wintel environment -but the wrong things to do with Linux because these are solutions to problems Linux doesn't have.
From a Unix perspective Microsoft client-server looks like something Rube Goldberg might have coughed up during a drug overdose but to the typical MCSE this is reality - and so if you make him use Linux, this is what he'll use it to implement. He won't do that because he's stupid, incompetent, dishonest, or lazy; he'll do it because he's an expert. You've heard the story about the frog and the scorpion? think scorpion.
Linux growth in Asia and Europe is going to be interesting to watch because of this. Overall, both regions have less MCSE style expertise but stronger heirarchial controls. For most, therefore, the change from Wintel to Lintel will be a change in supplier, not a change in architecture --undoubtedly why some companies in Europe seem to be exploiting national resentment against Microsoft, against the United States, and against the use of English just to power through a change in desktop OS supplier as a short term means of selling support services for Linux in client-server deployments. The result, however, should be increasing desktop costs in business and more emphasis on playphones (combination phones, cameras, PDAs, and gameboys) among consumers uninterested in the complexities of home PC use.
The Asian situation is differently driven but the forces at work may produce much the same outcome because client-server is client server whether implemented with wintel or lintel - the gains come from getting rid of it, not from changing suppliers for it. In China a unique combination of paranoia, political information control needs, and the drive for national economic independence are behind a focus on government possession and control of the source for any software of significant economic or military importance. That, in turn, makes Linux a natural for countries like Thailand and the Koreas where its adoption lets them keep pace with their oldest enemy (and biggest trading partner) while saving a few bucks, forging partnerships with big American firms like Sun, and reducing their WTO liabilities on IP and other copyright infringements.
Both cases represent important losses for Microsoft while creating commercial opportunities for Linux suppliers, but neither represents either a technical win for Linux or the beginning of a serious revolt against client-server. The likelihood, instead, is that both will lead to the kind of extensive commercialisation that becomes a barrier to further technical progress - dead ending Linux worldwide just as it enters its prime.