Hi BillM,
I don't know your background but I do feel some prejudices coming through
Is this for a research or hobby initiative (learn something, prove something) or is this for a commercial, sellable product?
This is an error that is often made. Commercial is not the opposite of 'open source'. The opposite of commercial is non-profit, amateur, hobbyists. The opposite of open source is closed source, proprietary. Projects can be any of the four combinations e.g. you can have non-profit closed source ones where the design is kept secret or you can have commercial open source projects (OpenSPARC). It is true that business models that work for one type of project most likely won't work for other types of projects. One of the boundary conditions in the EDA world is that validating IP in silicon is a costly business.
I am convinced though that currently a lot of man power is wasted for protecting one's own flow: iPDK/Python vs Cadence/skill, UPF vs CPF, oasis vs gds, ... A lot of man power could be redirected to real innovation if all could work on a common open ground. The latter should not necassary be open source but open to alternative open source implementations; which is not really the case for OpenAccess.
You need to understand how something was developed, validated and hopefully supported before you decide to integrate into a product that you intend to sell.
I think this works in favor for open source as there you can for 99% of the time see the history of the project and verify the design yourself as you have the source. For closed IP this is not always the case.
what terms must our company agree to to access Open Source (HW or SW)?
One key one: if I develop anything around or using this, do I have to provide this invention back to the Open Source group that the original Open Source HW or SW?
I agree that terms and conditions are very important. How many providers give any liability for their IP ? Most contracts I have seen start with the no-liability clauses so that the customer is on it's own if the design does not work as advertised; and there is no difference between open or closed IMO. I agree it all comes down to trust. So I see three possible choices for IP
- Do everything yourself. You are the owner afterwards and have full control of it. Of course this will cost a lot of man power to complete it.
- Use external open source IP. Hopefully this allows to reduce the man power needed significantly but may also force some license obligation like making your changes/optimizations public. You also need to trust the IP but you could do the same verification as you would do in the first point.
- Buy external closed IP. Again this will reduce the man power needed, probably more then for the previous option but part of the saved money will go to the IP vendor. You also need to trust your IP provider and need to follow the terms and conditions of the IP provider.
Personally I think the second option is too often excluded a priori based on prejudices as you have given in your original mail. I do know that part of chip design these days is failure risk management and it is understandable that project managers are conservative.
if there is an issue, where do I get support? is this 24x7 dedicated support or is this a loosely defined network of part timers that have 'day jobs'?
This same argument has been used by Microsoft and the UNIX vendors to try to stop Red Hat and Linux; certainly on the server (and in the EDA world
) this has not succeeded. I do remember some years ago that development at Mentor was done on Linux servers but no Linux version was sold as the marketing guys were afraid of it.
There is nothing stopping people to provide commercial support for open source; finding a good business model and the seed money is always difficult.
greets,
Staf.
