A friend of mine used to say, “You know how you kill a weed? You grow it to death.”

And buried down in the bowels of a Register Story about who might buy Skype, is this indicator that Skype might be learning a thing or two about growing itself to death:

In terms of technology, Skype has a real problem: it relies on “supernodes” - users who have direct Web access to a “real” IP address. The traffic in and out of normal nodes wouldn’t be capable of travelling between two subscribers; there are no inbound routes. So the software fakes a session through a supernode.

The problem seems to be: the number of potential supernodes is dropping, and the number of ordinary nodes - behind mapped addresses or firewalls, or both - is going up rapidly.

The result: quality of calls is falling. Bandwidth available is poor compared with a year ago.

This would be consistent with what we’ve seen in some informal testing I’ve been involved in. We thought it was related to quality issues in the implementation of their authenticated proxy support (and there definitely seem to be some issues there) or the overall load on our proxies, but perhaps we’ve been wrong.

If this is the case, then it looks like Skype’s free bandwidth lunch may be coming to an end. I see no reason why it wouldn’t be the case–I can now buy a Cable/DSL firewall/access point/print server/blender/kitchen sink combo at the grocery store right next to the extension cords and lightbulbs. Even the least computer-savvy people I know have bought and installed one. Okay…maybe I strongly encouraged them, especially if we were drinking, but the point is still that they all went out and did it!

But getting back to Skype… IP Bandwidth is cheap these days. I would think it would be easy for Skype to buy some IP access scattered around the country, set up some supernodes, and let the network do the rest. Unless they don’t have the cash, which seems highly unlikely considering how little this would cost since all they’re looking to do is augment the existing infrastructure.

This reminds me a little of the origins of Amazon.com. They were founded as an on-line bookseller who could beat the competition because they didn’t have to support a physical infrastructure of supply chain–they’d just order books from the publishers or dealers, then ship them along to the customer in return for some mark-up. It didn’t quite work out that way, though. These days, Amazon has massive “distribution centers” filled with inventory, workers, and fulfullment systems, none of them free.

I think that if anyone is working on a valuation for Skype, they should look long and hard at the assumptions about the viability of the pure peer-to-peer architecture and how much it’s going to cost to prop that architecture up as the supernode-to-non-supernode ratio continues to shift the wrong direction.