[cap-talk] Logging membrane in KeyKOS

[Norman Hardy]

I did implement it (ANNALYST), indeed in PL/I.
It could have but did not 'fake' memory segments.
I do not now recall why I didn't. It was not hard.
I used it to examine some of the software tools and found and fixed  
some performance problems therein.
You could not fake meters and meter references across the membrane  
went unmediated.

There were just a few closely held keys to kernel objects, (domain  
tool at least), that could not be fooled by fake arguments.
I think the last bug that I fixed in the kernel was discovered trying  
to put a membrane between the nodes of a domain.
Like the quarks of a hadron, the nodes of a domain could not be  
separated--even with the bug fix.
The domain tool required a real, not fake, node key as an argument.
It was fooled by a gate key trying to pass itself off as a node key.
It the fake node led the kernel down a buggy path and it crashed.

There is some explanation of the difficulty of virtualizing (faking)  
kernel objects here:
http://cap-lore.com/CapTheory/KK/Synthetic.html

I have not been keeping up on this list and Google notified me of a  
Keykos reference at
(http://www.nabble.com/Logging-membrane-in-KeyKOS-t4206880.html).
Thinking that that was some other mail list I answered there.
It seems to be some sort of mirror of this list.
Naturally my first reply there did not get to this list.
This is thus my 2nd reply.


On 2007 Aug 2, at 8:58 AM, Charles Landau wrote:

> At 4:40 PM +0100 8/2/07, David Hopwood wrote:
>> The KeyKOS documentation is not well organised, but it has many
>> hidden gems. I hadn't realised before that a logging membrane had
>> actually been implemented for KeyKOS:
>>
>> <http://www.agorics.com/Library/KeyKos/Gnosis/103.html>
>
> It was designed but I don't believe it was implemented.
>
>> The most significant limitation seems to have been that in KeyKOS
>> memory keys cannot be synthesized ("faked").
>
> Note the qualifier "Currently" with "the segment keys may not be used
> to define memory". I think memory keys can be proxied with a little
> work. (If you are proxying across a network it could be a lot of
> work.) This aspect was even less implemented than the rest of the
> design.
> _______________________________________________
> cap-talk mailing list
> cap-talk at mail.eros-os.org
> http://www.eros-os.org/mailman/listinfo/cap-talk


Norm Hardy: <http://cap-lore.com>

What has always made the State a hell on earth has been precisely  
mankind's attempt to turn it into a paradise.
Friedrich Hölderlin (1770-1843)