[cap-talk] Selling capabilities programming

[James A. Donald]

James A. Donald:
 > > The trouble with any capabilities language is that
 > > we are *not* going to rewrite the world's software
 > > into that language, thus capabilities languages
 > > cannot address the present crisis.

Sandro Magi:
 > To be honest, I used to buy that argument, but I don't
 > find it that persuasive anymore. The more expressive
 > and the safer your language, the more easily you can
 > rewrite existing software. MarkM demonstrated this
 > with his implementation of the Doughnut Lab. OCaml,
 > Ruby and Python have been growing in popularity and
 > supplanting existing software in C, C++ and Java.

If one wants to get a program up in a hurry, sure, you
can write it a lot quicker in Java, but if there are two
competing programs, one written in Java and one in C++,
pretty soon most people will be using the one written in
C++.  For example, I used to use Azureus, but now I use
μtorrent.

Further, in an environment where most programs have
dangerously great authority, the end user has little
incentive to demand a few programs with limited
authority.  It is only worthwhile worrying about
limiting authority if you can radically reduce the
amount of software on your system with great authority.

Absent critical mass, users will not demand, so
programmers will not supply.   If only a few programs
demanded excessive authority, they would come under
pressure, but when almost all programs demand excessive
authority, no pressure.