[DNSOP] Hostname (was: Comments on draft-ietf-dnsop-qname-minimisation)

["Niall O'Reilly" <[e-mail]>]

  Thanks, Paul.

At Thu, 08 Jan 2015 06:06:10 -0800,
Paul Vixie wrote:
> 
> yup. and true to my words, that url will work in some places, and not
> in others. as a local-seeming name it will often be tried with search
> lists first, which might find a dk.vix.com or whatever locally, and it
> will often not be tried as a dns name,

  Indeed, but that's not what I'm getting at.

> since the definition of a "host" in rfc 952
> and elsewhere is label . label [. label] ...".

  The relevant specification in 952 (at
  http://tools.ietf.org/html/rfc952) seems to be

      <hname> ::= <name>*["."<name>]

  which, IIUC, corresponds to "label [. label] ...".  After some
  moments' doubt, I noticed that other nearby use of the "*" notation
  is consistent only with an interpretation of "zero or more instances
  of the following construct" as, for example, in

      <protocol list> ::= <protocol spec> *["," <protocol spec>]

  As for "and elsewhere", I've followed the "Obsoleted/Updated By"
  reference tree from 952 (1123, 1349, 2181, 5321, 5966, 2474, 2535,
  and onwards), and can't find where a restriction to a minimum of two
  labels for a hostname is introduced.

  I'll be glad to learn that I've missed something.

> i am never cheerful about things that work differently depending on
> what OS, what libraries, what browser, what ISP, or what search lists
> and/or what dns content is "more local" to me.

  Me neither. 8-)

  Niall O'Reilly