Subject: How to run an FSB R&D lab (maybe)
From: Tom Lord <>
Date: Thu, 21 Mar 2002 00:34:08 -0800 (PST)

One of the most effective organizational structures for software
innovation that I've ever seen is the college campus.  There you have
a lot of people with substantial free time, intellectual curiosity,
and no particular job requirements.  The number of people is large
enough and general mixing of people effective enough that little
cliques form, break-apart, and re-form as interest rises and wanes in
various topics and projects.  There's plenty of machine and reference
resources around to work with.  There's a constant barrage of
stimulating lectures and the latest news of recent innovations
elsewhere.  There's a walled-garden network for communication and
results aggregation.  Projects that "make sense" gain momentum all by
themselves simply because people want to work on them.  Projects that
don't work out get dropped like rocks.  Influences come from far and
wind -- including from far outside the domains of techie geekdom.  The
pace is casual when there's not much going on, frenetic when it pays
to be frenetic.  People have fun and express their joy by doing
quality work.

That's where we get Gtk, from Berkeley: it came out of an
undergraduate computing club.  At CMU, in the 80s, before the web,
there was a sort of campus-wide web microcosm built out of AFS and the
Andrew Project software: I could rattle off a half dozen very
successful spontaneous projects, taking place in the space of maybe
3-4 years, that you've never heard of but that gave that campus, at
that time, very effective tools.  It wasn't even the case of a huge
hacker community out of which a half-dozen projects succeeded: there
was a lot of overlap in those projects -- it was a small community
being extraordinarily productive.

It's an _extremely_ efficient model: there's no proposal/approval
process.  There's no per-project budgeting pains.  There's no mental
gymnastics associated with trying to explain goals that make sense to
decision makers not prepared to understand them.  There's no pressure
to get a win out of a project that the participants have just about
given up on.  There's just a society of mind and spontaneous action
driven by informal consensus and excitement about newly discovered

So I guess the right model is to pay people to hang around, have a
budget for honorariums for lectures, buy stack privileges at some
nearby campus libraries, encourage people to set up comfortable
hacking nests, throw some good social events, patronize the local
establishments of bleeding edge culture, hire a mix of people as if
you were planning a really good party, then walk around a lot to keep
up with what's going on and figure out where you need to send in the
technology transfer troops.

Having shouted so long about pocket-protector process engineering and
serious new business models, I thought I ought to mention the other
side of the coin.


"Oh for the love of g-d!  Where's the ROI, man?!?!  WHERE'S THE ROI?!?"