[AG-TECH] AG IPv6 support

[Ivan R. Judson]

Both of these replies are very encouraging, we're very interested in this
work, and I know there are some Asia Pacific sites that are also interest in
IPv6 support.

I've stuck some comments inline:

> We (Southampton) and UCL worked on IPv6 for Globus Toolkit 2, 
> over a year
> ago, and had a significant part of it working.   However we 
> dropped this
> effort when GT3 came out, and have since done work to 
> port/validate GT3
> for IPv6.  This is now working, and includes JDK1.5 testing.   This is
> IP version independent, not a port to IPv6-only.

This is the work I believe I was referring to; I've chatted with Piers
O'Hanlan and Peter Kirstein about the IPv6 work. Piers pointed me at this
other work as well:

http://www.rd.center.osaka-u.ac.jp/globus/
 
> The vic and rat tools are both IPv6 capable, and have been 
> for a number of years (both were developed at UCL under the 
> MICE project, I believe). 

Colin confirmed these are in fact compatibile, and our vic is a derivative
of UCL vic.

> UCL and Southampton have some funding in the 6NET project 
> which could be used
> to demonstrate IPv6-enabled AccessGrid.   We both have 
> existing IPv4 AG 
> deployments.  We both have IPv6 multicast deployed in a 
> production quality environment on 6NET (www.6net.org); both 
> PIM-SM and SSM IPv6 applications.
> We have had ~30 partner vic/rat IPv6 PIM-SM sessions.   We 
> have used FLUTE
> and DVTS for IPv6 SSM.

This is what we are interested in understanding and enabling, can you point
me at more details?

> It seems to me that the GT2-based AG components are the main 
> problem.  We
> may be able to resurrect some of the old GT2 IPv6 work we 
> did.   Other bits
> may be more problematic.

Yes, this is currently the lump of work that I think needs to be understood.
 
> What would be useful would be a complete architecture 
> documentation for an AG node/system.  Could you point us at 
> such a document for analysis?

We have various documents of this flavor on the web (and in cvs):
http://www-unix.mcs.anl.gov/fl/research/accessgrid/documentation/index.html

And http://fl-cvs.mcs.anl.gov/viewcvs/viewcvs.cgi/AccessGrid/doc/

We're more than happy to answer questions as well.
 
> We should definitely discuss this in more detail once that 
> analysis is done, as we could be of mutual help :)

Let's plan on it. Are you all coming to the retreat? It'd be a great place
to sit down and map out a solution.

--Ivan