Comment With all due respect to Alan Cox, (Score 0, Troll) 323
It would probably be fair to say that none of us would be talking about open source or any of this stuff without him. I suppose there's always the perennial complaint about the "monolithic kernel" and the bazaar method of development, but as ESR effectively argues in The Cathedral and the Bazaar, the highly polished engineering represented by the Cathedral approach comes at the price of a snail like pace of development.
In today's technology market, rapid development is essentially to short AND long term success. Now while Alan Cox might have liked a better API for hardware drivers or an architecture more in line with the "microkernel" concept so popular amongst academics, it's just not the way to succeed in the open source world. This is what has made the .NET platform so popular: Its focus on rapid development using the latest in aspect oriented programming technologies makes it an indispensible business asset for shops with relatively small programming teams.
But I digress. The issue here is overengineering. It's been shown that creating working kernel based on a register machine like most modern microprocessors is NP hard. What this means is that there is an inherent degree of complexity that cannot be swept under the rug via APIs and microkernels or any of the other academic fads that come and go year by year. You have to make a choice somewhere along the line and Linus chose a path with a proven history, with tremendous success I might add.
In short, it's time for people to stop the backbiting and recognize that the monolithic kernel, ugly though it may be, is the best we got and we should be thankful for it.
In today's technology market, rapid development is essentially to short AND long term success. Now while Alan Cox might have liked a better API for hardware drivers or an architecture more in line with the "microkernel" concept so popular amongst academics, it's just not the way to succeed in the open source world. This is what has made the
But I digress. The issue here is overengineering. It's been shown that creating working kernel based on a register machine like most modern microprocessors is NP hard. What this means is that there is an inherent degree of complexity that cannot be swept under the rug via APIs and microkernels or any of the other academic fads that come and go year by year. You have to make a choice somewhere along the line and Linus chose a path with a proven history, with tremendous success I might add.
In short, it's time for people to stop the backbiting and recognize that the monolithic kernel, ugly though it may be, is the best we got and we should be thankful for it.
