The Internet Sky Is Not Falling

Last week our esteemed colleague from Nemertes Research Johna Till Johnson wrote that the Internet is in deep peril because: "IP itself is nearing end-of-life, with no ready alternative." To prevent what she describes as a looming crisis, Johna champions a radical new Internet architecture. Just as the sky did not fall on Chicken Little, the Internet sky is not about to fall on us.

Route scalability and address depletion problems do not yet constitute a crisis, and folks responsible for Internet infrastructure will continue to take steps to keep things running (more on this in upcoming posts).  Measured technical and political means will get us where we need to go--redesigning the Internet from scratch will not.

In Boston last week we attended the Pouzin Society organizational meetings and the FutureNet Conference Beer and Pizza "Shootout": Will Route Scalability and Address Depletion Create "The Perfect Storm?"  The consistent drumbeat from Johna and other presenters was that the Internet can't continue in its present form. Internet pioneer Mike O'Dell told Shootout attendees: "We are getting what we wrought, but we wrought the wrong thing," and John Day told the crowd that, "what was once good enough has led us through the woods to a dead end."

We can debate the gravity and urgency of the Internet's scaling problems, but the solution is certainly not the mid-flight wing and engine change proposed by the Pouzin Society.  In our view the Internet will evolve to accommodate growth as it has until now. The architecture proposed by John Day is so radically different it is inconceivable it could be built in our lifetime.  For starters it requires replacing or upgrading every one of the billions of endstations running the TCP/IP protocol stack. Although the proposed architecture is not fully defined, the part presented in the May 4th meeting has serious issues. It replaces addresses with names for applications and locations.  To avoid scaling limits, there is no central registry for these names.  All networks will be built by ISPs that can provide different services and different trust levels. 

You will be forced to trust that your ISP will send you to the real Google rather than its own version of Google.  Despite the fact that Pouzin Society members express distaste for walled gardens, the architecture actually encourages them.  The architecture is so diffuse and unstructured that communication paths will be a created from a long series of rendezvous and discoveries before actually sending data to a destination.  One attendee remarked that this could degenerate into something akin to an elaborate Italian banking transaction requiring multiple transactions before you can deposit money.

But let us assume that the John Day architecture sorts this out and becomes a viable alternative to today's Internet.  The Pouzin Society is in no hurry to get there.  The project needs an architecture description, protocol specifications, a development toolkit, experimental implementations, system evaluations, and revisions to get it all right.  The group is keen on getting it right lest they repeat mistakes of the past.  Mike O'Dell repeatedly warned against an early public demo of the technology because in his view, "The Arpanet was a demo that got accepted as a working system before it was finished.  We don't want to make that mistake again."

The bottom line is that even if there is an impending scaling crisis as Johna foretells, the John Day architecture will not be in place soon enough.  That aside, the architecture will have its own growth issues that are simply not yet well understood. 

Why bet on an unproven radical change when reasonable modest changes will do the job? 

We will propose such a modest set of changes in upcoming posts. 

• address depletion
• FutureNet
• Nemertes Research
• Pouzin Society
• route scalability