- From: Roy T. Fielding <fielding@apache.org>
- Date: Fri, 28 Feb 2003 17:12:11 -0800
- To: Peter Koch <pk@TechFak.Uni-Bielefeld.DE>
- Cc: uri@w3.org
>> Yep. Theoretically speaking it matches the way BIND works, but I >> agree >> that it is better for parsers as >> >>> qualified = *( "." domainlabel ) [ "." toplabel "." ] >> >> Added to the list as 038-qualified. Thanks. > > In the DNS the trailing '.' is never part of the domain name. It is > only > used in zone file format to explicitly declare a domain name as FQDN. Actually, it is used anywhere that a resolver is used. I have personally used it in the past to differentiate between an ill-fated College of Medicine subdomain "com" from the TLD com, for both e-mail and URI addressing. > Also, it's currently true that TLD names do not start with digits, > there's > nothing in 1123 (or 952) that would forbid this. Until 1123 was > published, > a digit was not allowed as the first character of any label, but this > was > what 1123 relaxed. The only remaining restriction in this direction is > that > a hostname must not "look like" an IP (v4) address, i.e. it would be > unwise to have a TLD consisting of digits only. That was not the intent of 1123. It allows all-numeric domain names only because it is well-known that TLDs will never be allowed to be all-numeric. It is therefore useful (and encouraged by 1034 and 1123) that we syntactically distinguish IP addresses from domain names. > I'd like to suggest that a hostname be either an FQDN (identified as > such > by having at least one dot) or consist only of a single label. > Everything > else is too dependent on the actual DNS searchlist strategy. Sorry, there are no implementations that correspond to such a restriction. ....Roy
Received on Friday, 28 February 2003 20:46:06 UTC