Underscores in host names

Thu May 19 15:53:53 UTC 2005

At 12:01 -0300 5/19/05, MARLON BORBA wrote:
>Hmm, they've always teached to me that . (dot) at the end of hostnames
>indicates the (hidden) Root domain:
>
>blah.domain.com.[Root]
>
>And my teachers always said that we don't need to write the final . because
>every domain belongs to the Root domain.

This thread needs to consider the layering of applications.  E.g.,

  Applications:   Mail, Web, things using gethostbyname    (host names, etc.)
--------------------------------------------------------
  Infrastructure: DNS, etc, (such as X.500)                (domain names)

Apps deal with host names, DNS deals with domain names.  To the DNS, 
host names are a subset of domain names.  To applications, domain 
names per DNS are "behind that interface."  Apps deal with host names 
and other names, all of which, if running over DNS, are mapped into 
domain names.

Referring to the above text, yes, a full qualified domain name ends 
in a "dot."  Whenever talking to or among DNS elements, the "dot" 
terminates the FQDN.  It must syntactically - a zero length label is 
the "null termination" of the domain name.

Applications passing domain names to the DNS (not strings of domain 
names) must have this terminating null.  However, this does not mean 
that the applications have to make the user or GUI require the null, 
as that interface is likely to be dealing with a string version of 
the name.

Some DNS applications, such as dig, don't require a dot at the end of 
the name - if one is missing, the dot is appended.  The user is 
alleviated of the burden of adding the dot - but on the other hand - 
the dot is forced upon the user.

Some applications, being agnostic, won't add anything to the user's 
input.  (This is true for non-DNS applications too.)  This gives the 
knowledgeable user more power - unique things can be done.  But it 
means that novices have more to learn.

Some applications assume the user is lazy and adds the dot in all 
instances.  Knowledgeable users get burned because now here are two 
terminating dots at times - until these users remember to "fall back" 
to not terminating domain names.

I've seen all three kinds of applications.  The latter ones tend to 
be the quick and dirty prototypes that don't see the light of day.

The moral is that applications are choosing when to add the 
terminating dot.  It's always there in DNS, but people don't access 
the DNS without going through an application.

As far as whether an "_" is "legal" in a host name, you can attack 
the question in many ways.

When cast into a domain name, "_" is legal.  DNS assigns no special 
meaning to that character in domain names.  Not even in the SRV 
record - if one reads the document carefully, there is no special 
meaning assigned by DNS, only a convention proffered that uses the 
"_".  The convention is a function of the applications using the SRV, 
not the manipulation of the SRV within the DNS.

When cast into a host name, "_" is not among the legal characters 
specified in the ancient documents.  But documents are just documents 
- arguing over what's legal according to them is about as useful as 
watching haircuts.  (Unless you are on a software design & 
implementation team.)

When thrown into applications, "_" may or may not have a special 
meaning.  Some applications will raise exception to "_".  The more 
interesting question is why?  Is this simply because of the ancient 
documents' restrictions?  Is there some parsing consideration?  Is 
there a special semantic inferred?

It really isn't important whether the character is "legal" or not 
(until there is a network police force).  What is more important is 
whether the character will work in all environments in which the name 
is needed.

Will the "_" work in DNS domain names?  Yes, unless there is a but in 
the DNS implementation (always a caveat).  Will the "_" work in a 
host name?  Only if the applications in use, referring to the host, 
can handle such a name.

-- 
-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
Edward Lewis                                                +1-571-434-5468
NeuStar

If you knew what I was thinking, you'd understand what I was saying.