Skip to main content

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [List Home]
RE: [hyades-dev] Agent to Agent communication

Sorry for the response lag on this -- I've been on vacation.

It is my intention for agents to be able to send any command across this
connection that they would be able to send to each other through the
HCE.  The only point of this is to eliminate the middle-man and thus
make communication a bit more efficient.  I'll make that explicit in the
document.

I assume named pipes are available on all the platforms we want to
support as that it the mechanism by which agents attach to the RAC in
the current implementation (and to the HCE in the new proposed
protocol).  There may be some workaround on one of the platforms I'm not
aware of.

As to what those platforms are, I would say the same set as Hyades
supports today, plus 64-bit Windows and Linux.  I suppose this is one of
the requirements we need to get documented.

-Andy

-----Original Message-----
From: hyades-dev-admin@xxxxxxxxxxx [mailto:hyades-dev-admin@xxxxxxxxxxx]
On Behalf Of Allan K Pratt
Sent: Monday, August 23, 2004 5:51 PM
To: hyades-dev@xxxxxxxxxxx
Subject: Re: [hyades-dev] Agent to Agent communication

What do agents do when they have a connection to each other? Can they
send 
each other all the commands and responses in this protocol?

I think sockets are common to all platforms of interest. Are named pipes

universal enough?

What are the platforms of interest, anyway? The same set as Hyades
today?

-- Allan Pratt, apratt@xxxxxxxxxx
Rational software division of IBM




"Kaylor, Andrew" <andrew.kaylor@xxxxxxxxx> 
Sent by: hyades-dev-admin@xxxxxxxxxxx
08/20/2004 04:26 PM
Please respond to
hyades-dev


To
<hyades-dev@xxxxxxxxxxx>
cc

Subject
[hyades-dev] Agent to Agent communication






Attached is a fragment describing two possible ways to establish direct 
communications between two agents.
 
I?ve described two ways: one using named pipes and one using sockets.
The 
named pipe has the advantage that it would allow the agent to use the
same 
mechanism to process commands coming directly from other agents that it 
uses when the commands are coming through the HCE.  The disadvantage of 
the pipe connection is that it is not secure.  The socket connection 
allows us to add security, and it will be reusable in the future if we 
ever want to communicate directly with agents on other machines.  The 
disadvantage of the socket connection is that it is a heavier solution.
 
Take a look at this and let me know what you think.
 
-Andy


Back to the top