[AG-TECH] Well - multi-node DOES work, but ...

[Ivan R. Judson]

You didn't do anything wrong. The assumptions you made we are working
towards enabling, but as you said the current tools are of a certain age,
and the effort to author new tools is something that is difficult to say the
least. The existing tools (while a bit aged) have uncountable hours of use,
introducing new tools *will* cause some unease since they'll introduce new
bugs, etc.

That being said, we're looking at having some things in place in the 2.3
release (scheduled for later this year) that *might* make what you're
describing go away. We have two very specific cases that we need to address,
rat and vic (sending video). We've found solutions but have to put some work
into them to make this work. We're doing some of that work this software
release; you'll notice node service bundles now package their respective
binaries with the package. This (while introducing some difficulties) should
make it so we can more easily bundle and ship node service packages and they
don't rely on anything outside of the package.

This is key work for being able to package up new (ahem, ui-less) node
services, without colliding with existing ones.

Anyhow, long story short, we think as of this release, folks can go about
their merry way creating all sorts of new node services and that the toolkit
should support them being plugged in and tested in a very clean way. 

--Ivan

> -----Original Message-----
> From: owner-ag-tech at mcs.anl.gov 
> [mailto:owner-ag-tech at mcs.anl.gov] On Behalf Of Randy Groves
> Sent: Sunday, April 18, 2004 8:16 PM
> To: ag-tech at mcs.anl.gov
> Subject: [AG-TECH] Well - multi-node DOES work, but ...
> 
> Following Lewis Grantham's notes, it looks like I 
> successfully got my two machine node to start up.  But I was 
> disappointed to see the RAT and capture vic interfaces start 
> up on the 'subordinate' machine.  With no obvious way to 
> control them from the 'control' machine.
> 
> But what is the point of having a venue client on one machine 
> that can't, for instance, tweak the rat or vic on the other 
> machines?  You still have to have windows open to both 
> machines.  A small amount of thinking about this does make me 
> realize that my assumption was wildly improbable - I mean we 
> are working with tools of a 'certain age'.  Not to say 
> they're OLD, but ...
> 
> Is a development like a venue client that can truly control a 
> multi-machine node (all functionality concentrated in the one 
> client).  I'm guessing that this may require non-gui versions 
> of the tools.  Or something new.
> 
> -randy
> 
> Or did I do something wrong?
> 
> 
>