[cap-talk] Some good advice for disruptive technologies

[David Barbour]

On Mon, Jan 10, 2011 at 9:12 AM, Karp, Alan H <alan.karp at hp.com> wrote:
> I haven't been promoting a killer application as much as a killer use pattern, simple service chaining.

I hope you're having success!

Platform technologies are difficult to sell because no
individual wishes to absorb the costs of driving the
adoption, nor the risks of being the only to adopt.

Further, unless you catch the audience in the narrow
window where they are exploring architectural design,
they'll be reluctant to change simply because they
are invested in their current architecture.

So there is a lot of difficulty in selling an architectural
'pattern', and that's even assuming the audience was
relatively security conscious and interested, which may
be a tad optimistic.

A concrete 'killer app' (or service), however, would be
something that could sell on its own and that could carry
caps to anyone wanting to integrate with it.  While James
might dislike the notion of 'exposing' capabilities, doing
so would actually be a calculated decision: one goal of
the killer app would, after all, be to sell capabilities by
making their use obvious, clear, and necessary for any
integration.

The real problem is finding a killer application that cannot
be easily performed without capabilities. I.e. the goal would
be that any clones also use capabilities (since that would
also be a 'mission accomplished' ending).

My own interest is in developing a distributed programming
language and plugin extensible VM with powerful support
for automated code distribution (automatic sharding between
client and server, mobile agents, compute cloud integration),
but I also have several ideas for killer applications atop the
system that leverage the security properties of capabilities
and other properties of my language (reactivity, eventual
consistency). Even a language can't sell itself.