ftp://xcf.berkeley.edu/pub/src/local/viola/violaIntro.ps.gz
Marc VanHeyningen <mvanheyn@cs.indiana.edu> wrote:
> - A secure language for which secure implementations are "easy"
The viola language system is "object-oriented" (not pure in the Smalltalk
sense, but has big object units...). This makes it easy to enforce safety
by a policy of marking imported objects as untrusted. As such, those
imported objects' scripts have no system priviledges, no sub interpreters,
can't coarse other objects to do things in a dangerous way...
> - A reasonably simple language that can be learned and implemented in
> a vaguely concise way
It's somewhat C like in syntax, has very simple grammar. It's typeless,
or rather, has automatic conversion and has internal data types (this
has efficiency implications).
> - A language which allows a variety of presentation approaches (e.g. a
> single program can use a GUI interface or a simpler
> interface, either through abstract routines or more directly
The viola system is more GUI centric. But the abstraction could be
knocked up higher such that the GUIs might be more sensibly replaced
by other interfaces (ASCII, voice, etc).
> - Portable
The system has almost no Unix or X ism. Should port very well.
> - Extensible
The language currently doesn't allow dynamic sub-classing.
> - A language with a secure reference implementation which may be used
> freely by developers (i.e. it is unencumbered by things like GPL or
> other licenses that may inhibit development)
> - Reasonably efficient
The script is compiled to bytecode before interpertation.
> and so on. I don't know of anything that comes closer to meeting
> these requirements today, or shows promise for doing so in the near
> future, than Safe-Tcl. You can use Safe-Tcl to retrieve a program
> from a server and run WWW-oriented code on it today. I
> We can argue all day about how "it should be scheme-like" or "it
> should be more object-oriented" but unless some effort at actually
> developing the thing is underway NOW, it's a non-starter. We want
> this to actually be working and in widespread use in a reasonably
> small number of years.
The violaWWW development on this front has been underway for a while now.
Needs refinement, but it's basically already working in ViolaWWW,
in the WWW context, such that applets may be built and linked/embedded
into toolbar or documents.
There's more info in the paper on this.
-Pei Pei Y. Wei
O'Reilly & Associates, Inc.