[Numpy-discussion] grant proposal for core scientific Python projects (rejected)

Joe Harrington jh at physics.ucf.edu
Thu Apr 18 15:47:53 EDT 2019

Hi Ralf,

The rejection is disappointing, for sure.  Some good ammo for next time 
might be the recommendations in this report from the US National 
Academies of Science, Engineering, and Medicine:



You can download a free PDF if you click around and give them an email 
address.  There is some code on the cover that might raise a smile.

Lorena Barba, Kelle Cruz, and many other community members contributed, 
both as committee members and as white-paper authors.  While it's a 
report for NASA, the conclusions are strong and there is explicit 
support of investment in community resources like numpy, scipy, astropy, 
matplotlib, etc.  Other agencies are asking similar questions, so I 
expect the report to get somewhat of a look at NSF, etc.

A note on the Academies' process: A consensus study report must only 
include statements that no single member of the committee objects to, 
and the committee generally only includes senior members or people with 
a lot of relevant experience.  There were stakeholders from some large 
modeling shops for whom openness might be a threat to their business 
model, in their eyes, and some of the senior members had never 
experienced the open-source environment and had bad experiences sharing 
software.  This made for an interesting social dynamic, and prevented an 
all-out recommendation to forcibly open everything immediately.  Given 
all that, I was ultimately pleased that we got full agreement on the 
recommendations we did make.

So, some general thoughts on fundraising, not specific to this proposal:

1. Try NASA.  The Administrator for Space Science, Thomas Zurbuchen, is 
pushing "open" very hard, given the success of open data in NASA Earth 
Science, and its positive impact on the economy in fields like 
agriculture and weather forecasting.  He paid for the study above.  Many 
grant programs specifically solicit proposals for open-source tools.  
There are also technology development programs in other parts of NASA 
than the Science Mission Directorate.  Try contacting Dr. Michael New, 
who is Zurbuchen's deputy, and could direct you to appropriate programs. 
(Please, let's be coordinated and not all deluge the guy.)

2. As suggested in another message, it's often easier to get support for 
a specific, targeted item as part of a big project or institute using 
that item, such as LSST or the black-hole group. There's a certain way 
to wend into those projects, usually originating from within.  STScI has 
long devoted programmer resources, for example.

3. There's such a thing as a share-in-savings contract at NASA, in which 
you calculate a savings, such as from avoided costs of licensing IDL or 
Matlab, and say you'll develop a replacement for that product that costs 
less, in exchange for a portion of the savings.  These are rare and few 
people know about them, but one presenter to the committee did discuss 
them and thought they'd be appropriate.  I've always felt that we could 
get a chunk of change this way, and was surprised to find that the 
approach exists and has a name.  About 3 of 4 people I talk to at NASA 
have no idea this even exists, though, and I haven't pursued it to its 
logical end to see if it's viable.

4. I mostly lurk here, since being more actively involved in the early 
days of numpy docs, so maybe this one  has been tried already, or is in 
the works.  My apologies if so.

Think of development as a product to buy.  You could put chunks of 
development up for sale, advertise them, and coordinate one or more 
groups buying them together.  For something like an efficiency boost, 
you could price it according to the avoided cost of CPU resources for a 
project of a given size (e.g., somewhat below the net present value of 
avoided future AWS cycles for the projects buying it).  It would be like 
buying a custom-built data pipeline, except that once you buy it, 
everyone gets it.  This might mean scoping out a roadmap of 
improvements, packaging them into fundable projects with teams ready to 
go, pricing them, advertising them to specific customers and in trade 
media and shows, and making sales pitches.

This sounds really weird to us scientists, but it would work just like a 
regular purchase for services, which the government and industry are 
much more used to doing than donations to open-source projects.

Don't just sell what the customer is buying, sell in the manner that the 
customer likes to buy.

5. And, keep trying grant proposals to NSF!


On 4/18/19 6:36 AM, Ralf Gommers <ralf.gommers at gmail.com> wrote:
> A number of core projects (NumPy, SciPy, Matplotlib, Pandas, 
> scikit-learn) got together and put in a proposal to NSF for a large 5 
> year grant, and it was unfortunately just rejected. We now published 
> the proposal, which may be of interest: 
> https://figshare.com/articles/Mid-Scale_Research_Infrastructure_-_The_Scientific_Python_Ecosystem/8009441.

More information about the NumPy-Discussion mailing list