Licensing Questions Looming

Word on the street is that some software companies are considering charging more for licenses on dual-core-processor systems. They want to count each processor core as a whole processor instead of basing license prices on the number of processor sockets.

It's pretty obvious this will mean more money for software companies. What's not so obvious is that this may not be very fair. The cores still have to compete for bandwidth to communicate with the rest of the system, so a dual-core system does not necessarily provide the same performance as a traditional dual-processor system.

Processor manufacturers and software manufacturers seem to take one of two sides in the debate: They want licensing fees either based on the number of sockets or on the number of cores, respectively. A third possibility is a scheme somewhere in the middle--a technological three-fifths rule, where a dual-core processor would be considered not quite equivalent to two whole processors.

That logic places us on a very slippery slope. Once you start itemizing the processor in the cabinet to calculate licensing, it opens a Pandora's box, tempting us to look at the rest of the architecture. If software vendors want to price based on performance--the real reason behind processor-based pricing--then there's no end to the metrics they can use. For example, what clock speed is the processor? How much memory is used? What is the speed of the memory? And what about those hard drives: Do they run at 10,000 or 15,000 rpms?

Sponsored post

What's more, software vendors would acquire numerous details about how many of each type of system you build or sell each month. That begs another question: Should license fees increase each time you perform an upgrade?

I'm stretching the point to make a point: It's one thing to charge fees based on the type of system used--uniprocessor vs. SMP--it's quite another to license based on the parts in those systems. If you think likewise, let software vendors know that now. Once a licensing model is in place, it's hard to uproot, and per-core licensing cuts into the bottom line.

Which side do you take? Let me know via e-mail at [email protected].