Re: A bit of bomb throwing....

From: David Karger <karger_at_mit.edu>
Date: Wed, 18 Jan 2006 11:09:58 -0500

I've seen several mentions in this thread of the idea that while the
goal of the open source community is to build stuff (clearly a sign of
brains and productivity), the goal of academics is to publish papers
(clearly some vain form of social parasitism). As a parasite---er,
academic---I assert that in fact both communities have the same goal,
which is to make things better than they are now. We simply have
different horizons. Open source is about the software that you can
build right now. Academia is about the software that you might be able
to build 5 or 10 years from now. The reason we write papers is that, by
definition, the thing we are describing can't really be built right
now. While research invariably involves building prototypes (such as my
group's Haystack client), those prototypes are typically too fragile to
deploy for broad use, because they are trying to do something that we
don't fully understand how to do yet.

Some may argue that if you can't build it for real now then there's no
point in working on it at all, but I would say this is a lot like
driving at night with your headlights off---sure you can see 2 feet in
front of the car, but is that really sufficient? You need to try out
ideas to see them fail, so you can fix them, and it would really slow
progess down if we only tried out ideas that were practical right now.
Under the current research system, by the time something becomes
practical, researchers have (hopefully) played around enough with it in
its impractical stages to let us know how to build good solutions
quickly. Just to snag one example that appeared on this thread, Timbl's
creation of the web didn't just happen; it arose from a background of
plenty of previous research on hypertext that did not result in broadly
deployed systems, but that did produce a background of ideas that others
could draw upon to build new artifacts.

So yes, academics may suffer enough human vanity to be disappointed when
their "pet algorithm is discarded in favor of something better for the
product", but that is irrelevant to the broader goal of getting that
algorithm out there so that it can be picked up and used 5 years from
now when it is just the right thing for the next product.

David Karger
Professor, EECS
MIT Computer Science and AI Laboratory
32 Vassar St.
Cambridge, MA 02138

Michael McDougall wrote:
> Zack Rosen wrote:
>> On Jan 17, 2006, at 1:44 PM, Prokopp, Christian wrote:
>>
>>> It sounds like you are fighting windmills. The modern economy works the
>>> way it does (early adopter problems etc.) and the university have
>>> totally different interest than open source communities, companies or
>>> you and me. No one will change that (soon) for good reasons - it is not
>>> perfect but the best we know of.
>>
>> I have yet to hear an explanation for why academic research should
>> not be applied towards real world problems in partnership with
>> open-source communities other than 'this is just not the way things
>> are done in academia'. Can someone please lay out the reasoning for
>> me?
>>
>
> I thought I made it pretty clear.
>
> - Different goals: academics want publishable advances in science,
> while open source communities want to create usable tools. Quite
> often, you don't need scientific advances to create usable tools, you
> just need good engineering and design, which is not at all the same
> thing. So an academic may work on the project just to see their
> (scientifically legitimate) pet algorithm discarded in favor of
> something better for the product. Or conversely, the open-source
> developers may feel like they have to include complex technology that
> users don't care about just to keep the scientists happy. Open-source
> communities generate code, academics generate papers.
>
> - Coordinating with an outside group brings in extra overhead and
> extra risk. I mentioned a couple scenarios above. You need to monitor
> the mailing lists, and track all the changes that might impact your
> research. If something needs to be implemented to fit with your
> research project you have to convince other people to merge that into
> the project. Like I mentioned above, it's possible that the community
> will decide to move the project in a direction that's bad for your
> research--what do you do then? What if the head of the open-source
> community starts trashing your funding agency in the national media?
>
> I've seen research projects destroyed by these kinds of issues. It's
> much easier for a researcher to create a small environment where they
> own the code, they know all the people involved, and everyone is
> working towards the same end (publishing papers).
>
> I'm not saying academics should always hide from the world. If their
> research project has common goals with some open-source community,
> then maybe the extra overhead and risk is outweighed by the benefits
> of real-world experience and users. Certainly, I would rather see
> academics extend existing projects than duplicate existing tools--and
> this seems to happening more often (there's a ton of academic projects
> making Eclipse plugins, for example). However, I understand that it's
> not always feasible for an academic to align their research interests
> with the interests of some open-source community.
>
> Michael
>
> Michael
Received on Wed Jan 18 2006 - 16:09:38 EST