Timothy Normand Miller wrote:
I specifically grilled them about this and had the director of Technology
Transfer explain to me the boundaries carefully. If I'm not using any of
BU's resources, they don't lay any claim to it. And actually, it's not
really BU; it's the state of New York, since this is a state institution.
Basically, if I'm doing this from home, using my own equipment, it's the
same as if I did outside consulting under the same terms. I DO need to
disclose potential conflicts of interest, and I have to keep some kind of
paper trail that shows that my contributions were published not using BU
resources. That's one of the reasons I want to post code to the list using
my gmail account.
You have a pretty generous IP agreement (good for you).
Why don't you spend ~$1k and run your employment contracts by a lawyer
specializing in such things (work for hire). I have been told that New
York law is in general pretty funny compared to the other states; I don't
know how that might effect you.
Do you know how much I make? Let's just say that I took a big pay cut when
coming to BU, so this isn't an expense I can make right now.
Guess ~100k. Even if I were 40% off (not happening) I'm pretty sure you
could do that without having to eat beans for 6 months.
SHOULD I let NYS own this?
Short answer: yes
I think you should do what is best for your career. Neglecting what has
already been created, what you create going forward (apparently) is
_your_ ip and nobody has any title over that unless you choose to
utilize the university resources. Looking over Binghamton's website
leads me to believe that they are trying to build up research programs.
I don't know what opportunities exist or CAN exist if you develop that
IP under the umbrella of the university. Ga Tech has enormous resources
for commercializing ideas in a variety of ways; you should have some of
the same there. If you can pick up the phone and talk to a patent lawyer
paid by the university _that really means something_. Yeah you say the
state takes 60%, but look at the risks they take, and look at the risks
that you no longer personally take.
What kind of asset do you want to be to your new university home? Do
you want to help them build a program as part of a long lasting
institution, or are you just passing through? I'd bet that the choices
you make now will be remembered when you come up for tenure.
I could probably license this under BSD or MIT, and no one in NYS would
raise an eyebrow. It it would prevent _us_ from getting any revenue;
meanwhile, Apple could plop it into the iPhone 12 and make billions. I
don't like that. I'm doing this for research purposes so that academics
can have an easier time developing new GPU-related techologies, not so that
some money-grubbing corporation can just lift our IP.
#1 What you create going forward is going to be basically all yours;
certainly given your copyright assignment. Perhaps you want the
publicity of a/the 'project' which good for the school and you. I
certainly don't have any problem with that. I cannot fathom any random
mailing list dude having any serious form of title or say in the usage
of any funding that you get from what you will develop. You can maintain
your stated academic goals by releasing in GPL with the copyright owned
by the university.
#2 Just how many people on the mailing list are going to labor to
develop then assign copyright of contributed code to you personally?
Certainly not I, though I wish you no ill will. I don't care anything
about revenue. I read and post to the list as I suspect most others do
for general interest. As an aside, I am going to finish my tvc sort of
as I described it some months ago (under the public domain) purely for
that same general interest.
#3 Just what sort of revenue stream are you expecting (don't answer)? I
punch 3d graphics core into google and I get dozens of hits. The
Intels, googles and apples of the world have zero trouble getting
anything they need. Is selling a single license to somebody as an
individual worth not building a stronger relationship where you are?
#4 Are you sure that the GPU as classically imagined and discussed on
this list is interesting to CS research folks; in other words have
people moved on? Are you so sure that grossly brute force (but easy to
use) solutions like the Xeon phi are not going to 'win'? I'll allow
that in the deep embedded world probably not.
Basically I'm arguing that _if_ you keep it totally internal with an
external GPL license, you loose little to no proficient collaborators,
you gain the ability to build your relationship with your university,
you gain access to the assets that the university can provide, you take
fewer personal risks, and you maintain your stated academic goals. If
you then license the design to some random outfit the hardware is
already conveniently documented to open software standards and there may
be a bit of community developed (L)GPL software floating around that can
talk to it. Certainly keep the mailing list going as a discussion place
and for community access. Heck if other people on the mailing list want
to scratch an itch and develop hardware that integrates with something
you have designed, they can, under the GPL.
Kinda mini rant:
This is _way_ outside of my field, but as I stare into my crystal ball,
all of this foss graphics driver bs is going to go away over the next
5-10 years regardless of what you choose to do. I think that amd,
intel, and nvidia are all moving to have their parallel processing
elements able to work in the same address space as your program.
Basically you will make a function call in your own address space,
protected by the mmu and you access all the power of the 'gpu' elements
for general computing (that use their own local fast memory, but mapped
into your address space). I believe this means that programming of the
elements is going to be completely documented. For intel, that job is
'easy' for the Xeon phi. After that point, composing a display is just
blending memory buffers together, exactly what the wayland people are
talking about. Mesa would turn back into what it was originally, a
software rendering library -> I think that mesa + LLVM are effectively
the first steps in that direction.
Yeah, I realize that some custom or optimized bit of hardware may be
able to to it x% faster with y% less power, but hey vhs beat betamax and
odds are we are mostly looking at x86 machines.
-John
--
John R. Culp
[email protected]
_______________________________________________
Open-graphics mailing list
[email protected]
http://lists.duskglow.com/mailman/listinfo/open-graphics
List service provided by Duskglow Consulting, LLC (www.duskglow.com)
