A plea to PSI and other clueless DNS Admins

[Mark_Andrews at isc.org]

> I understand underscore in domain names is not RFC compliant.  Is there any
> particular reason why it is not allowed?  It is allowed in SRV records etc. a
> s
> you stated.  Just curious.

	A number of reasons.

	1. Conversion between characters sets.  Not all platforms
	   use ASCII internally, this is still true today, and you
	   needed to be able to convert between character sets
	   cleanly in both directions.  A subset of ASCII that
	   worked well with these character sets was choosen.
	   Underscore is a character that is known *not* to convert
	   well.

	2. Some characters are just bad to have as there can be
	   confusion about whether they are there on not in certain
	   contexts.  Underscore is one of these.  Try printing
	   out a line underlined which contains underscores.  This
	   is also why tabs and spaces are considered the same in
	   may contexts.

	SRV records deliberately choose a name syntax that would
	not clash with hostnames and hence with delegations.  The
	issues above should not be a problem with SRV records as
	lookup routines should be taking the name without the first
	two labels, service and transport and constructing the
	query.

	Mark

> 
> /stanley
> stanley.liu at tmca.com.au
> 
> Cricket Liu wrote:
> 
> > > Oh but wait until people start sticking Microsoft ActiveDirectory servers
> > on
> > > the Internet.  They are loaded with underscores in AD and Microsoft says
> > > that the underscores are RFC compliant.
> >
> > Actually, they're not yet RFC-compliant, but they are ID (Internet Draft)
> > compiant.  Read RFC 2052bis, at
> > ftp://ftp.is.co.za/internet-drafts/draft-ietf-dnsind-rfc2052bis-02.txt.  Th
> e
> > key here is that underscores are only
> > forbidden in domain names that are used as host names.  The owner name
> > in a SRV record is a service identifier.
> >
> > > Ofcourse they also say that stub
> > > records are NOT RFC compliant, hence they won't include them in thier
> > > implementation of DNS.
> >
> > I believe you mean stub zones, and again, I think they're right.  Stub zone
> s
> > have been labelled experimental for as long as I can remember, even
> > though they're very useful.  And I've never seen an RFC that formalizes
> > their definition or use.
> >
> > cricket
> >
> > Acme Byte & Wire
> > cricket at acmebw.com
> > www.acmebw.com
> >
> > Attend our next DNS and BIND class!  See
> > www.acmebw.com/training.htm for the
> > schedule and to register for upcoming
> > classes.
> 
> 
--
Mark Andrews, Internet Software Consortium
1 Seymour St., Dundas Valley, NSW 2117, Australia
PHONE: +61 2 9871 4742                 INTERNET: marka at isc.org