Iljitsch van Beijnum writes:
I guess not because I have no idea what you're talking about.
There is a natural tendency to think that by dividing a 128-bit address
field into two 64-bit fields, the address space is cut in half (or
perhaps not diminished at all). However, in reality, dividing the field
in this way may reduce the address space by a factor of as much as
nineteen orders of magnitude. Again and again, engineers make this
mistake, and render large parts of an address space unusable through
careless, bit-wise allocation of addresses in advance.
For example, if you assign addresses sequentially, you can address up to
2^128 hosts with a 128-bit address. If you divide the address field
into two 64-bit fields--one for the country and one for those within the
country, say--intuitively you might think that you've only slightly
reduced the total number of available addresses. In fact, however,
you've dramatically diminished the size of the address space. Unless
you have exactly 2^64 countries in the world, each of which contains
exactly 2^64 hosts, very large segments of the address space will be
wasted. The effective address space may be many trillions of times
smaller than you expect. And you may find yourself exhausting that
address space far more quickly than you ever thought possible.
When you divided an address field into bit-wise fields, you encode
information into the address. This introduces redundancy, which reduces
the usable address space. If you allocate a country code to one country
in the example above, for instance, and that country has only 1000
computers, you've just wasted roughly 18,446,744,073,700,000,000 machine
addresses. Do this for multiple countries, and you could easily waste
almost all your address space simply due to this extremely foolish
division of the address field. It's unlikely you'll have 2^64 countries
to accommodate; and it's equally unlikely that each of these countries
will have exactly 2^64 hosts (no more, no less) to address, so you are
wasting many bits of the address field. And the wasted space increases
exponentially with the number of wasted bits. Wasting ten bits instead
of five bits doesn't waste twice the space, it wastes _32 times_ the
space.
I've often felt that the most common mistake made by engineers in
designing systems, especially computer systems, is underestimation of
capacity, and wastage of existing capacity. Dividing address fields
into preallocated zones is the most common example of this. They just
never learn, and they cannot be told. They are always convinced that
there's more space than anyone will ever need in their address field,
and by the time they find out otherwise, it's too late. It happened
with IPv4, 16-bit addresses, and 32-bit addresses; it will soon happen
with 64-bit addresses and with IPv6. I don't understand how engineers
can be so stupid in this one specific domain.