A-label definition

John C Klensin klensin at jck.com
Tue Jun 24 13:31:24 CEST 2008



--On Tuesday, 24 June, 2008 07:10 +0200 Frank Ellermann
<hmdmhdfmhdjmzdtjmzdtzktdkztdjz at gmail.com> wrote:

>...
> Right, we talked temporarily about different things.  For the
> at the moment existing www.example.com I'd sometimes say that
> www is the host name, and www.example.com the FQDN.  And at
> other times I'd say that this is host www.example.com, which
> is clearly inconsistent.  You had www.example.com (the FQDN)
> in mind, I had www (the label) in mind.
>...

And this is why I've tried (unsuccessfully) on several occasions
to simply eliminate "host name" from our vocabulary.   The term
is used in different places and by different people to identify
either 

	* the first label in an FQDN
	
	* the first label in an FQDN iff it identifies a host
	rather than something else.
	
	* an FQDN that identifies a host.

There are some similar issues with "domain":  Can it be a single
label or some combination of labels that is not a FQDN?  Does it
refer to an FQDN without a leaf node (e.g., a hostname-FQDN
without  the first label)?

Curiously, I was motivated to try to make disjoint { A-label,
U-label, LDH-label } definitions precisely because I saw us
going down the same road with terms like

	* punycode string (does it contain the prefix or not?
	Used both ways in various contexts)
	
	* IDN (a label or an FQDN?  Does it or does it not
	include all-ASCII, non-punycode-encoded and prefixed,
	strings?.  Used in all of those ways.)

IMO, whether the particular words chosen are right or need
tuning and regardless of how we tune the definitions, ending up
with three distinct and non-overlapping terms, possibly plus
some things like:

	* IDNA-invalid label: something that is none of a
	U-label, A-label, or LDH-label.
	
	* IDNA-domain: an FQDN that contains at least one
	U-label or A-label (i.e., whose labels are not all
	unprefixed LDH strings).  Note that even this is
	ambiguous, because it is unclear whether an IDNA-domain
	as defined that way would include SRV FQDNs with
	U-labels or A-labelsin the rightmost label positions.
	So, if we need this at all, we need to do better.

john




More information about the Idna-update mailing list