Really dont want to go into lessons on root server allocation and dns structure on the web but one has to remember that the dns server is not responsible for the handling of packets streams that pass through a network just tells them where to go.
By nature there are only allowed 13 root dns servers in our current dns structure. It is only through anycasting that allows for addition of new root servers. (ie servers that can share the same address to provide redundancy) As of 2007 I think there were only a total of 130 total root servers in existance. Known as A - (I dont remember what leters they are on now).
Root dns servers by default should not have to handle but a minimum of data exchange if the subordinate dns servers in the web update as they should.
Problem is as traffic increases on the web, more unknown quaries make it past the closest dns server to the root server. This is due to malformed data, unknown or incorrect data input, increased server replication, and denial of service attacks, to state only a few.
As this traffic increases it is only probable that more data will reach the root servers. You also have to take into consideration that a root server has to have a script installed to "translate" the character into its punycode. This requires processer usage decreasing the overall performance of the server itself.
Look at Poland. Currently only 30 to 40% +/- of the population is using the internet on a reguar basis. There are hundreds of reasons why. These usage stats are going to rise. If and when the idn "mainframe" is in place you should expect the usage by "native language" input to cause an exponential increase to an already increasing traffic load relative to the base rate of growth.
Poland by itself does not present a large challange but Take into consideration the 4 billion Chinese people that are going to be increasing usage and you can see why it is possible that data exchange at the dns root server level is going to increase with the release of billions of new idn domain names over the years.
Now looking at the disallowing of certain domain characters you have to look at their potential for use. Granted a few characters by themselves are nothing but try this
Say we want to keep @.com. No big deal by itself but then someone wants to reg
[email protected] then
[email protected] then
[email protected] and eventually someone says there is a huge market for 12 letter @ domains because all of the other domains are taken. (LLLL.com) The number of disallowed domains is huge.
254 place holders for a name server, all with a combination of @. Now add the ! then the ^ then the & then whatever and you can see the number of domains that will be disallowed is huge. Granted disallowing the characters may be more for cosmetics and to reducephishing than to reduce traffic, but he action itself is going to decrease the potential for domain registrations.
We have to stop thinking like English speaking natives. The majority of people in the world do not speak English and do not want to be forced to learn the language, no more than many English speaking people want to take the time to learn Chinese.
The internet is still in its infancy. Ask someone from the 70's when the internet was nothing more than a big old Sperry Remmington running a couple of discussion boards with a dial up connection. There is a long way to go. Traffic is going to increase. Population increases, Businesses go online every day. It just a matter of fact. You may not see and increase in ASCII people going to idns but you are going to see native speakers increase traffic to idn domains.
As always jmho