[cap-talk] Need Challenge Problems

[Jed at Webstart]

At 03:02 PM 7/19/2006, Karp, Alan H wrote:
>Jed wrote:
> >
> > In my opinion anything that can effectively work can work at
> > the network level.  At the network level it becomes quite clear what can
> > be done and what can't be done.  Once a scheme for communicating 
> permissions
> > can be agreed to and standardized at the network level, then I believe
> > it's fairly straight forward to implement such a scheme at the OS and
> > even I expect at the language level.  Trying to go in the other direction
> > (start at, say, the language level or the OS level and try to extend to
> > the network level) I belive is a prescription for disaster - as I 
> believe we
> > have seen repeated again and again.
>
>Client Utility started from within a machine and projected outward.
>That greatly simplified both our intra-machine infrastructure and
>application writing.  The main advantage was that there were only local
>references.  On the other hand, the intra-machine protocol was message
>based, so you could argue that we didn't really do what I claim.

My concern is that, as far as I know, Client Utility is effectively gone.
KeyKos is gone.  NLTSS is gone.  EROS is gone.  Amoeba is effectively
invisible/gone.  Even more recent efforts at capability architectures
(e.g. E, Joe-E) have at most very limited and isolate applicability
and a questionable future.

These systems are not only gone in the sense of not still running and
having no market share, but even the concepts they developed are
pretty much only historical anecdotes or seemingly academic exercises
discussed in places like cap-talk and e-lang.

By contrast consider a protocol like ftp.  One can debate how much it
contributed or what it's value is, but whatever it's value has been it's been
lasting and a force of standardization and binding between systems.  I
remember debating some of the DO, DON'T, WILL, WON'T aspects of ftp
negotiations in the middle 1970s, long before TCP.  The ftp protocol transited
to tcp and is still going.

Similarly with protocols like http and ssh, ntp and syslog, smtp, etc.
Once a protocol becomes established on "THE" network, it continues
to provide whatever value it delivers in a lasting way until supplanted
by something better (e.g. ssh replacing telnet to a large extent, though
even there that seemed to be more commercial than technical as telnet
could have easily been extended).

One can certainly argue that operating systems and languages can
make important and lasting contributions (e.g. consider Unix and Windows,
Fortran and C and Perl, maybe even COBOL - heh).  Architectures at
those levels can even have influences that spread to other areas like
dynamic library loading mechanisms or subroutine calling conventions.

Regardless, I still believe that a network object reference mechanism
(communicable object permission or network "capability") would have
the best chance of initiating a lasting value in the area of capability
architectures.  Any such architecture would provide an automatic
standard for use at the OS and language levels.  I believe it would
also greatly simplify the space and tighten the design constraints
on such communicable permission tokens at the OS and language
levels.

--Jed http://www.webstart.com/jed/