In message <39A8AABA(_dot_)C966BECB(_at_)ecal(_dot_)com>, John Stracke writes:
If "active network" means "the endpoints can program the
routers", then probably never. It's an interesting idea, but
it'd be much too hard to make it secure and scalable.
Funny, that's what I think too (and I worked in the area for 4 years :-)
At least, not with the building blocks we have now.
Also, it goes against the end-to-end principle: keep the routers
simple, make sure the network delivers packets freely, and let
the hosts implement about any more complicated behavior they
need. IP is all about smart hosts; AN is all about smart
routers. ATM was about smart routers (switches), too, and that's
(part of?) why it didn't take off: ATM switches had too much work
to do, so they didn't scale. But an AN router would potentially
be working much harder than an ATM switch.
That's not necessarily true; if a hypothetical active network infrastructure
was used to quickly deploy IPv6 on all the routers, there would be no
deviation from the end-to-end model, and it would arguably be a good thing.
It all depends on the flavor of active networking you're thinking of (data
vs control plane, modules vs packets, administrator vs user, etc.) There have
been some good survey papers on IEEE and ACM publications the past couple of
years, and the relevant DARPA page has a number of pointers to funded projects
(no, I don't remember the URL off the top of my head).
That said, your comment about complexity is all too true.
-Angelos