Re: [dnsext] does making names the same NEED protocol changes at all?

[Nicholas Weaver <[e-mail]>]

On Feb 24, 2011, at 9:17 AM, Phillip Hallam-Baker wrote:

> The problem here comes in that the downstream server needs to be prepared to support the scheme. And there can be multiple levels of ambiguity so the cache requirement grows exponentially with each level.

Why?  That would be case C:  A 0 TTL CNAME or DNAME is synthesized, so there is no support needed down the chain.


And again, caching and similar requirements do not grow exponentially IN PRACTICE:

How much bigger would your ASCII DNS cache be if you included every capitalization variant as a separate cache entry?  People don't TyPE liKE THis, but DNS supports people TyPEiNG lIKe thIS for ASCII.  

You'd end up with WWW.google.com, www.Google.COM, and the like, but you won't end up with wWw.goOgLe.cOM in the cache.



The problem only occurs if the downstream resolver has a different set of normalization rules, but developing and specifying such a standard is outside the purvue of DNS: we aren't the language experts.  But given a standard normalization, no resolver or protocol changes are needed, and only authorities which support such normalizations need to change.  Additionally, since the upsteam authorities do need to know the state of the downstream authorities, they can ensure that the normalization is agreed to.