As it now looks like IDNA will be defined as an RFC despite all the problems it will result in, I have looked at what it will mean for DNS and for my system. I will discuss why it will not be usable for me in an other mail and here just take the problems with DNS.
Of what I can see, IDNA will prevent a simple solution to DNS using native UCS! IDNA will force all solutions using native UCS in DNS to require the client to first test if the server can handle UCS and the do the real query. This will result in a big overhead in DNS lookups. While it can sometimes be cached, it cannot always. In my original ideas I used the IN bit in the header to allow the client to tell the server that it could handle UCS. It was also possible to use a flag bit using EDNS. With this flag, the server could respond correctely to both old and new clients, without any overhead. If IDNA is used, this is no longer possible. With IDNA the client must do a query that results in server failure for older software, because it must know if the server handles UCS or not before asking the query (is not needed if IDNA is not used). So you have to do two calls to the DNS server, if older software. And as many calls go through more than one DNS server, each step going from new to old server software will result in two calls. Is it acceptable that a native internationalisation of DNS will be much more difficult and have much more overhead because IDNA is introduced in some applications? The more I look at IDNA, the less usefull I see it. Dan
