The goal here is a working network. If every 192.* "C" network that had
been allocated in the long ago were to be advertised tomorrow, that would
add 2**16 routes to the global table and a lot of the net would fall apart.
If every wants-to-be-ISP got a /19, address space wastage would be immense
and we would be into the n>=224 "E" multicast space already, with the end
clearly in sight. Previously allocated blocks are not reclaimed when an
ISP goes out of business, they usually pass on with the technical folks
and they soon show up as part of some garage-band ISP elsewhere.
Market and technical pressures have established an equilibrium. It is a
damned shame that a group of people with lots of money and technical savvy
in the data communications field cannot just start up and compete head to
head with more established players, competing on the basis of price and
service levels and so on. Peering and address space have become barriers
to entry and this has been universally bad in the history of communications.
As the existing players discover the horizon effects on growth, such that
an ISP over a certain size can no longer simply grow in order to add
customers, they will start spinning things off rather than integrating them
vertically. That's the point where newcomers will next have an opportunity
to enter the market without severe barriers.
It is also dimly possible that ubiquitous ATM, and IPv6, and NIMROD will all
bear fruit and the market will enter a healthier period of total chaos. For
now the barriers to entry are real, and the people whose participation is
needed for changing them, are too busy growing, buying eachother, and making
tons of money to be bothered levelling out the playing field. We'd already
be reading about a Consent Decree if the problem weren't so international in