Your PC industry history is a little off.
Perhaps more than a little, thanks to your thoughtful presentation of the unfolding events at the beginnings of the PC. But I wonder if a full exposition of detailed, apparently accidental outcomes properly reflects the context in place at the time these momentous decisions occurred? I suspect individual personalities and seemingly random decisions were largely the inevitable consequence of techy-political turmoil in the early eighties, and by no means accidental or random outcomes.
The Microsoft and Intel empires were accidental. Wind River will never be another Microsoft. Even Microsoft will never be another Microsoft.
We agree that Microsoft was in the right spot at the right time, and benefited from IBM's decision to farm out the PC operating system. You imply this event will never be repeated because it was an accident, one dependent on so many unusual events, that the probability of a repeat is infinitesimal.
How do you explain this: In 1995, Intel, the dominant microprocessor company in the world realized that a new processor/operating system was needed in all the world's server (and perhaps peripherals that connect to servers) to implement a concept they called I2O. Intel named this processor the IOP, which they would manufacture, and the operating system, IRTOS. Intel realized that the IOP would succeed only if it became a standard, and therefore turned over responsibility for I2O specifications interface to a users group called the I2OSIG -- after involving the leading RTOS company to contribute to the IRTOS initial specifications and development a working instance which they called IxWorks. Although Intel had an in-house RTOS, called iRMX, it decided that they lacked the kind of expertise needed to structure the IRTOS appropriately, at without putting in more effort than it was worth. Intel also knew that the IOP only would succeed as a standard if the operating system was non-Intel. Otherwise there would be apparent conflicts anathema to potential participants important to the successful launching of the I2O standard. Finally, Intel wanted to make sure a full set of state-of-the-art development tools would exist for the IRTOS when it rolled out, and that mandated out-sourcing the operating system. As a consequence, Intel scrapped iRMX, selling to an embedded hardware company.
Clones of the IOP began to appear within a couple of years, first by Digital, then by Symbios Logic, both using ARM cores. Intel liked that because it encouraged I2O becoming a standard, and showed acceptance by the computer industry. Intel was comforted by the knowledge that its IOP dominated market share, and was the first choice of users concerned about compatibility.
The IOP ramped a little slower than originally hoped, mainly because Microsoft was a willing but strangely unenthusiastic. I wonder why? After all, at the time Microsoft's future depended on extending the Wintel computer up the enterprise, where the main barriers were I/O throughput, reliability, maintainability and clustering. The IOP addressed all these problem areas almost magically. By April 1998, for example, the TCP record was set with a RAID device using an IOP in the controller. Netware, SCO Unix jumped on board as quickly as possible, and by 1998 even some non-Intel Unix boxes began to capitulate. With Windows NT 5.0 committed to supporting I2O, by 1999 the lion's share of servers were I2O dynamos.
But the number of uses of I2O didn't stop at making servers faster and better. First it spread to virtually all devices that connect on LANS and communicate to servers, not so much to enhance speed, but to enable sophisticated network management schemes. BMC took the lead in the department by extending their Pilot product, which became an overnight sensation with companies managing large networks of computers. This lead to a demand for the IOP on NICs connected to each PC, which in turned forced Microsoft to issue a patch for all their Windows products to support I2O. It also forced, all Unix and even mainframe companies to become I2O compliant.
I2O blossomed way beyond the network. Semiconductor fabs started adding an I2O IOP core to their systems-on-a-chip, in response to demand, wherein it quickly became an absolute must for anything communicating to anything. While the ARM I2O core originally was most popular, in time I2O cores were developed by a number of SIP firms, driving the price of the cores to competitive levels. The Intel-compatible I2O IOP chip became a relic, but to this day Intel continues to produce tons of commodity I2O chips, both standalone and integrated with other cores.
As I2O IOPs proliferated, it became obvious that backward compatibility and the ability to upgrade and manage billions of such devices put a premium on the use of the highly-evolved IxWorks and development tool-set. Service companies formed that focused exclusively on I2O designs and IxWorks, like Serano Systems in Colorado. All these effects combined to confer total domination of the IRTOS space to IxWorks.
Of course at the time the momentous decision was made to involve the leading RTOS vendor, no one could foresee the importance of IxWorks. It was just a random event that can never be repeated because it was an accident, one dependent on so many unusual events, that the probability of a repeat is infinitesimal.
Allen
Epilogue: Not only does history repeat itself, but in this case, it already has. There will be another Microsoft. Not just by accident, but because Microsoft's are needed and they always will be. Events about I2O may not turn out exactly like I told them, but then again they might. And if they do, it wasn't an accident. |
