Complete I/O is the difference between a kernel and an operating system. Think of the total effort that goes into building, integrating, testing and documenting a complete set of I/O subsystems.

• File Systems (3-5 types of media)

• Networking ( 10-20 protocols)

• Serial communication buses ( 5 – 7 options)

• Graphics displays ( 3 or more options )

These I/O subsystems represents many years of work and many hundreds of thousands of dollars in investment if done properly. If you decide to undertake this task yourself, you must recognize the total cost of taking this approach.

• Savings can be created by providing a subset of features; however, porting is the approach you should take so really the feature set is determined by the ported software.

• Integration is 50% of the total development cost for a feature. If you decide to port software it will cost 5%-10% of the development cost to do the port but you still have to integrate it which makes the overall cost 55%-60% of the total cost.

• Documentation costs of 10%-15% of the total development cost are typical. These expenses are substantial because engineers and technical writers must be involved. You might throw this away or minimize it but if you don't create it, you will increase maintenance expenses.

Consider taking a kernel (why not a free one) and a bunch of I/O components and integrating them yourself. How long will it take to do this porting, integration, testing ( you will have to write this or just ignore it ) and documentation? Unison and DSPnano Open Source versions have a file system, serial I/O, kernel, testing and documentation off the shelf based on open standards. Its Free! How can you make money replicating this with proprietary kernels and undocumented open source software?

Another major issue is supporting new devices.  It is also expensive and time consuming.  By having an well defined I/O Framework, the ability to add new device support quickly and easily gives you Complete I/O.  This cannot be over emphasized.  While some RTOS suppliers offer no I/O model and others provide their specific brand of I/O, what you really require is a blend of your I/O and off the shelf I/O for most applications.  It is expensive to rebuild existing modules and it is also expensive to invent an entire I/O system - an I/O Framework provides the best of all worlds.

If you need commercial versions, the upgrade cost for the commercial versions will be returned to you the first time you need:

• Another component

• Networking with good performance

• Networking which is tailorable to the application

• Advanced network protocols

• Advanced kernel features

• To restructure the system

• To train someone

• Or customization help to meet a deadline or eliminate complexity.

Complete I/O based on open standards is the cornerstone of developing cost effective OEM products.