Re: Solving the right problems ...

On woensdag, aug 27, 2003, at 18:52 Europe/Amsterdam, jfcm wrote:

> Using a 1D numbering plan to support 5 (?) distinct dimensions (plan, 
> technology, network routing, global host addressing and interface 
> identification) looks like X121 or old telephony.
> Which additionnal information do you need to move accross if you use 
> fixed size dimension and gives evey dimension its separate numbering 
> authority/structure and then concatenate them into a multidimentional 
> address? Like for mobiles.
I have three problems with this.

First, "fixed size" always means "wrong size".

Second, with going from 20 to 40 bytes for the IP header I feel we've 
pretty much used up our overhead increasing budget for this decade.
Last but not least: having an identifier and a locator in the same 
packet is useless as we have mechanisms to authenticate identifiers and 
mechanisms to authenticate locators, but without mechanisms to 
authenticate the association between a locator and an identifier, we 
open up a whole new world of denial of service. Fixing this by carrying 
auth data (= a signature) inband would be an unacceptable increase of 
overhead, and doing it out of band would be redundant as it is much 
more efficient to carry the actual data out of band than the 
authentication data for the inband data.
> I remember when CCITT voted they would never allocate addresses of 
> more than X bytes. (was it not 32?).
32 bytes? This wasn't for voice, was it? I really wouldn't care much 
for dialing 75 digit phone numbers...
The way I see it:

We now have FQDNs that identify services or hosts, and we have IP 
addresses, that identify hosts or routing+interface. In order to 
decrease the ambiguity to managable levels, we need to move the FQDNs 
and especially IP addresses away from identifying hosts, and introduce 
and explicit host or "stack name" identifier. Host identifiers should 
look like IPv6 (maybe optionally IPv4) addresses, in order to be able 
to keep using our current protocols that expect 16/4 bit values. (And 
there isn't much of substance to be gained by giving them a different 
shape.) Where necessary, FQDNs can be replaced by application specific 
namespaces with a system to map those names to host identifiers to go 
along with such a new space. The hard part is coming up with a way to 
do the host/location mapping in a way that is simple, fast, cheap, 
secure, flexible and reliable.
Iljitsch