The Oracle OpenWorld conference is not going well for the database giant. Attendees were left wondering whether 2004 had ever happened after Larry Ellison virtually repeated his keynote speech from last time. That's an unusual way to celebrate nearly a year of Oracle Database 10g, the company's flagship product: the old Ellison would have been talking world domination by now.
The next keynote was even more embarrassing. Honoured guest Scott McNealy had missed the memo that said 'be nice to your hosts': he took the opportunity to lay into Oracle for not understanding dual-core processors. Why charge for two licences for one chip, he said, just because it has two cores?
He has a point. This is a fault line between hardware and software that will only widen. Traditionally, software companies have charged one licence fee per product copy: easy to understand and manage in a world where computers are simply single processors in unconnected boxes. But physics dictates that chip makers must now put multiple processors in single chips to remain competitive. That works fine for them: they can continue to cut costs and increase performance. Yet this advantage will be utterly lost if software companies promptly multiply licence fees.
Oracle is not alone in arguing that if two copies of its software are running simultaneously then two licence fees are payable, no matter what hardware is involved. This is not only short-sighted, it is self-defeating. Assuming a single core chip that runs twice as fast as a dual-core, both parts will run the same amount of software: why then should one attract twice the licence overhead? So far, this hasn't been a problem. There have always been faster chips.
That option is no longer available. By imposing a huge tax on the only sensible path for hardware innovation, software companies will either freeze the market or drive users to open-source software -- whose licence model is utterly agnostic about hardware configurations. Neither option will buy Ellison many more yachts. Moreover, as virtualisation and grid computing become facts of life, the disconnection between software and the hardware beneath will become complete. One piece of software may be running on any number of processors anywhere in the world: the old licensing model is as relevant to this as longbow practice is to the Marines.
If Oracle wants to be taken seriously as a player in this new world, it must reform its licensing and move on quickly. It cannot hold back the future by merely repeating its past.
