[Top] [All Lists]

Re: Inconsistency?

2005-04-17 19:42:01

John P Baker wrote:

In RFC2821, a clear definition of the structure of a domain
name is provided. In this definition, the length of a domain
name may not exceed 255 characters, and consists of one (1)
or more ?labels?

Two or more, because you have at least one dot, or you have a
domain literal without labels.

Each label must begin with an alphabetic character

| sub-domain = Let-dig [Ldh-str]

Letter _or_ digit.

In RFC2822, the definition is more obscure

RfC 2822 "allows" the absolute maximum before it definitely
breaks syntactically.  That's light years after you have lost
any plausible semantics.

OTOH you can see in RfC 2821 that narrow specifications of IPv6
domain literals also have their drawbacks, and the "general"
literal in RfC 2821 is much obscurer than the RfC 2822 syntax.
Ignoring many NO-WS-CTL in RfC 2822 for the moment, that was a
bad idea, as far as I'm concerned

a much broader subset of ASCII

Oh yes, incl. the funny NO-WS-CTL.  It took me several months
of whining, ranting, moaning, and outright threatening to get
rid of it in USEFOR Message-IDs.

?(_at_)? appear to be valid within a domain name.

It depends, DNS allows it, but it's not okay in the FQDN of a
host name used for mail (A or MX).

Does RFC2821 correctly reflect the syntactic requirements for
domain names, or is it too incorrect

Domains without dot (a.k.a. TLDs) are possible, try it with a
`host ai`: ai =  The author said that the required
dot in 2821 is an intentional feature, IMHO that makes sense.

I'm not sure about the IPv6 domain literals in RfC 2821, the
tag IPv6: is rather odd.  The General-address-literal is also
odd, and IIRC it's incompatible with IPvfuture in RfC 3986.

The latter is a STD, therefore it's "better" (but unfortunately
it has absolutely nothing to do with SMTP ;-)

                             Bye, Frank

<Prev in Thread] Current Thread [Next in Thread>