RE: Receiver of a signal

Subject: RE: Receiver of a signal
anders.ek#telelogic.se
Date: Thu Dec 28 2000 - 11:32:38 GMT

Dear Rick,

I think the intention was to create the instance sets statically, not
dynamically. The text changes that you propose seems like a good way of
making this clear.

Regards
Anders
-----Original Message-----
From: Rick Reed TSE [mailto:rickreed#tseng.co.uk]
Sent: den 22 december 2000 23:00
Cc: meeting
Subject: Re: Receiver of a signal

Dear Eckhardt, Thanks for the reply.

Dear All,

I think there is another issue that we have not considered related to the
need for routers that Eckhardt mentions:

If we have a create request using an agent type identifier and there is no
instance set of the indicated agent type in the agent containing instance
that performs the create, an instance set is created in the containing
agent. So there is a mechanism to dynamically create instance sets where
none existed before.

However, it seems to me that there are no channels leading to this instance
set are created.

To send a signal to the instance by TO requires the instance to be
"reachable": that is there must be a channel to it. Therefore it is not
possible to send signals to a specific dynamically created instance, even if
its Pid is known.

Aside: Actually and OUTPUT to a Pid that is unreachable is allowed, but the
signal is discarded.

The relevant text in 11.13.2 Model is:

"c) If there does not exist any instance set of the indicated agent type
in the containing agent then:
i) an implicit instance set of the given type with a unique name is
created in the containing agent; and
ii) the <underline{agent} identifier> in the <create request> is derived
syntax for this implicit instance set."

It is clear that "<underline{agent} identifier>" should be "<underline{agent
type} identifier>".

It would be possible to reword the text as static rather than dynamic as
follows:

"c) If there is no instance set of the indicated agent type in the
containing agent then:
i) an instance set of the given type with a unique name is implied in the
containing agent; and
ii) the <underline{agent type} identifier> in the <create request> is
derived syntax for this implicit instance set."

In this case the channels would be derived as normal provided the model
transformations are treated in the correct order which seems to be the case
in Annex F.

In Annex F this particular transformation does not seem to be mentioned in
the list at the end of F2. 

--
Rick Reed - rickreed#tseng.co.uk
Tel:+44 1455 55 96 55 Fax:+44 1455 55 96 58 Mob.:+44 7970 50 96 50

Eckhardt Holz at holz#informatik.hu-berlin.de  wrote on 22/12/2000 12:51:

> answer (d) is the right answer. I did some historical investigations
> on some older versions of SDL-2000 and found that we actually did have
> initially answer (a) as the right answer. This was back in time when
> agents where still caled "active entities". There were some rules
> stating 
> that a signal arriving at an agent is delivered to the agents state
> machine
> in case there is a channel for this signal to the state machine,
> otherwise
> an appropriate channel is selected and the signal is delivered to a
> contained
> agent.This does  of course require some kind of signal distribution
> inside an
> agent (router).
> These rules were removed/restricted with the introduction of typed PIDs
> and of the
> implicit interface.(cf TDT 614 and TDT 620 of the 1999 Toulouse
> meeting).
> The basic idea now is: If you adress something by a PId or typed PId,
> then
> you do mena the state machine of the agent with that PId.
> 
> On the other side in Oslo we had a short discussion on
> additional adressing schemes (with gates and interfaces) so may be we
> should 
> look into this, whether this could fulfill your intuition.

This archive was generated by hypermail 2a23 : Mon May 05 2008 - 20:30:55 GMT