>

[Brian Boles at Microchip wrote]
>      Only I2C implementations in SILICON (custom or ASIC) are subject to
>      royalty.  Users of that silicon or software implementations do not
>      have to pay any royalty.

My impression, going along with this, is as follows:

[1] Use of any licensed I2C device in a system conveys a license to use I2C
    throughout that system.

[2] Producing a slave device, whether entirely in software or in custom
    hardware, may require royalties if that slave device is used in a
    system with no licensed I2C devices.  Note that certain CPU's are
    considered to be licensed I2C devices while others are not.  I am
    also not certain if all I2C-compatible memory devices are licensed;
    I would not be surprised if some are not--see below.

[3] The addressing, start, and stop sequences used in I2C are not in any
    real way "novel"; the things that are novel are the "clock holding"
    and arbitration methods.  As a result, I would not be surprised if
    use of I2C in a system with only one master whose slaves never need
    to hold the clock is permissible, but use of I2C in a more complex
    system required either the use of at least some licensed hardware or
    else the payment of I2C royalties; sales of an I2C-compatible _DEVICE_
    may be viewd even more strictly.

[4] Brian made mention of custom silicon or PLD's; what about use of
    standard logic for I2C handling?  I think a clock-extender and start-
    detector could both be built using only a pair of quad open-collector
    2-input NAND gates.