I spent last week at a Dagstuhl seminar on
permissioned blockchain (thanks Mic for getting me the
invitation!). If you aren’t aware what this is, it is
basically an unstructured week at a German castle in the
middle of nowhere talking about research problems in a
particular area of computer science (in this case,
permissioned blockchain). So, basically it’s nerd camp for
adults—your mileage may vary, but I thought it was really fun.
As you all might expect, Hyperledger came
up quite a bit in discussions. In fact, Fabric seemed to be
the most talked about (and built upon) system. I had no idea
that there were this many people across the world in academia
working on things related to Hyperledger (it seems to be the
case that blockchain papers are being sent to a very wide
variety of conferences so it is hard to follow). In
particular, Hyperledger has captured a lot of interest in the
database community which I did not expect or know about until
last week. The fact that Hyperledger has caught on in parts
of the academic community was really encouraging. Some of the
academic work included direct building on Hyperledger (like
the fast Fabric paper—one of the authors was there), while
other work used, say, Fabric, as a way to test the performance
of new algorithms. For instance, multiple people reported BFT
algorithm tests in terms of Fabric performance.
However, there were some notable issues:
pretty much all of the participants didn’t know how to
contribute their work back to Hyperledger! Those that had
contacted people found the contribution process difficult,
thought it was hard to get started, and didn’t know who to
talk to about issues in the process. Many of these people
were not just coming with algorithms on pencil and paper—they
had modified versions of, say, Fabric running with their
implementation changes, and performance numbers to boot!
Several groups said that they tried to get involved and
contribute, but one or more hurdles stopped the process.
It struck me as particularly wasteful that
we did not have an efficient way to get these folks involved
in Hyperledger. Given that many of these research groups
already had working code, it seemed like it should be easy to
incorporate these changes, but it wasn’t happening. In
particular, I think this was due to the fact that most people
had never worked with an open source organization before and
were not aware of how things worked.
With this in mind, I’d like to suggest we
create a forum for Hyperledger research-related activities.
I’m not sure whether this should be a working group, SIG, or
something else entirely, but I think we should have a biweekly
(or perhaps monthly) meeting where researchers could talk
about their work and get feedback on how to contribute the
results of their research back into the Hyperledger code
bases. In addition to helping researchers contribute code, we
could potentially do more: we could have engineers talk about
interesting problems they face that might be good for
research, and researchers present solutions to problems (or
efficiency/security improvements) that could be implemented to
improve the various Hyperledger projects.
I asked people at the seminar if they would
be interested in joining something like this, and roughly half
of the 30 participants expressed interest in joining. So I
think we would have a pretty substantial crowd.
What do people think about this? Does
anyone have any suggestions on how to best implement this
idea? Again, it seems very wasteful not to help these
researchers (and potential contributors) get involved.
Thanks a lot for your time, and have a