On Fri, Jun 9, 2017 at 2:54 AM, Roman Treutlein <[email protected]>
wrote:

> Okay so make Mind-Agents managed their own resources. Yea that makes sense.
> You'd still need some simple Agent-Manger but yea I can see how that would
> clean up the overall design.
>

well, there are also some deep theoretical problems with the naive
conception of "mind-agents".  I suppose I should write this in a wiki
somewhere.

--  its surely just a lot easier to just launch a thread if you need that.

-- the naive idea is that mindagents are scheduled, but how? this is a
really hard problem that OS designers struggle with. What if a mind-agent
does file or network i/o? etc. naive conceptions of ecan can't solve this.

-- for robustness and reliability, its usually better to split up large
things into multiple network servers, so that if one small component
crashes, it doesn't take down the whole system.  Jamming all mind-agents
into one giant address space goes counter to this.

-- a classic example of where mutiple adress spaces are viewed as
importnat/good is for data applications: you don't want a rogue mind-agent
corrupting the atomspace. classic examples include: bad apps can't crash an
OS kernel (recall, with MSDOS/windows 3.1, Mac os9, they could, hello, the
birth of linux), bad database apps can't corrupt or crash databases (in
dbase/foxdb, they could; oracle started as a tiny company cause they had a
solution to this data corruption problem, called SQL. Ingres turned into
postgres).

adress space management is how this is accomplished.  we should not regress
to having rogue / wild / beta / untested mindagents running in the same
address space as the atomspace, where we may have some semi-precious data
we are trying to keep in order.

Even humans (mammals) have something called the blood-brain barrier, meant
to keep dangerous crap in your blood for getting into your brain. to keep
you alive/alert in case you get viruses in your blood.

So this naive conception of mindagents has to be completely reformulated.

--linas

-- 
You received this message because you are subscribed to the Google Groups 
"opencog" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To post to this group, send email to [email protected].
Visit this group at https://groups.google.com/group/opencog.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/opencog/CAHrUA35geHBRvER1a3UGO9dTVh7VO3LD5ySyya2r5QVRE1egqA%40mail.gmail.com.
For more options, visit https://groups.google.com/d/optout.

Reply via email to