[cap-talk] Capabilities in Cal/TSS

[David-Sarah Hopwood]

Charles Forsyth wrote:
>Norm Hardy wrote:
>> I believe that in Cal/TSS the sending of a message to another  
>> 'process' designated the recipient by an unprotected global integer.
>> This is a fatal design in the modern capabilities view.
> 
> event channels (queues) were accessed through capabilities.
> 
> http://www.mcjones.org/CalTSS/CalTSS.pdf points to a good handful of papers and
> reports about the system.  i particularly liked the more abstract IBM report
> "The Control Structure of an Operating System", which had a big influence on me when it first appeared.
> i remember writing a heavily stripped-down subset once in B, to get the feel of it.

Hmm. <http://www.mcjones.org/CalTSS/archive/691010-overview-bwl.pdf>
describes it as a C-list system, with one working C-list per process.
The description there seems to be inconsistent with message targets having
been designated by unprotected global integers.

Norm, maybe you're referring  to the "Master Object Table" index? If I
understand correctly, that seems to have been an implementation detail
not visible to user processes. There are also "class codes", but they
seem to be protected capabilities, from the description given.

-- 
David-Sarah Hopwood