This is a great thread, and something that I've had to wrestle with a lot.
Thanks to T&C for their insights - big fans...

This is a tremendous question, isn't it?  Deserving of a book in itself.
Sure, there's Adaweb at the Walker (I hope, haven't checked for a couple 
years).  But how long will we be able to access it, or even be able to see 
it?  It's a big question.
Jon's Variable Media solution is a strong strategy, and one of the positions 
I generally support.
I also believe that some archival of New Media should be attempted, although 
I often feel it too unwieldy in a long-term sense (I'd like to see New Media 
curatorial staff consider lifespans of 50-100 years as a general thought 
experiment and 'crazy goal').  In many cses, I realize this is unreasonable.

However, I do like the idea of 'graceful decay' - really lovely.  In many 
ways, it forgrounds the concept of the art, and less dependence on the media 
used.  Really strong idea, and one I hadn't thought of.

On my part, here are three things that I use as rules of thumb currently for 
a personal philosophy of New Media.
1: Time imparts performativity as well as ephemerality, and therefore works 
well with a performance model.  From this acceptance, emphasis is on 
documentation rather than archival.  It is less than satisfying, but 
presents a very stable strategy for long-term access to the basic 
information.  books are not going anywhere, and is why I am printing out all 
of my email for the last 5 years in 2008.

2: I have begun creating New Media on static platforms, i.e. embedded 
processors, dead game platforms, small electronics platforms, or derivative 
from very 'old' tech.  This is not all of my work, though (for 
clarification)
These works, if networked, create their own nets, and are not dependent on 
infrastructure.  Secondly, static platforms often have no moving parts, use 
communications protocols which have been in use for many years, require no 
software upgrades, and in general are 'plug & play:", allowng for easier 
archival.    .

3: If possible, also use formal components which are not dependent on one 
medium alone.  This fits with the 'graceful decay' model, and fits with the 
gesamtkunstwerk idea to an extent.

But I think this idea of considering public New Media art is a great thing 
to think about?

I would like to ask how our participants feel about architectural New Media 
sites, like the Tokyo Prada building which ncorporates live projection, the 
site (that evades my mind) in Germany which incorporates the display on the 
entire side of the building, and so on...

The question is whether incorporation of computational elements and New 
Media in architecture can compromise their lifespan?  This is a nasty 
question.