[comp.sys.hp] Gripe about nodename restrictions

bentson@grieg.CS.ColoState.Edu (Randolph Bentson) (08/17/89)

I've a copy of RFC 1034 (Domain Concepts and Facilities) and have
checked therein before starting this mild flame...

As you may notice from my address, we have a three part domain
name following the node name. It seems to work well here and
we see no reason to change.  However, HP seems to think that
one should only have a two level domain name, e.g. HP.COM.  When
I issue the command "nodename anon.cs.colostate.edu", I get a
message "nodename: invalid node name syntax".

Back to RFC1034... check section 3.1 last paragraph and the example in
3.5 for mention of a three part domain and size limits. Also note
RFC1035 sections 2.3.4 and 3.1 for more on size limits.

I suspect that a really aggresive competitor could disqualify HP as a
bidder for failing to meet this spec.  Has this happened?  (Remember
DEC getting bumped from the DOD procurement a year or so ago, based on
their failure to provide the "standard" Unix?)

Yet more flaming is based on the hostname length restrictions.
While this isn't as well founded as the other part, it's still
irritating to come upon limits (of eight characters), seemingly
based on tradition.

stevem@pserv.UUCP (Steve Mestad) (08/18/89)

In article <2425@ccncsu.ColoState.EDU> bentson@grieg.CS.ColoState.Edu (Randolph Bentson) writes:
>                                      HP seems to think that
>one should only have a two level domain name, e.g. HP.COM.  When
>I issue the command "nodename anon.cs.colostate.edu", I get a
>message "nodename: invalid node name syntax".

I think the command you want is domainname not nodename.  Nodename is a 
command that is part of HP's Network Services.  Domain name is part of HP's
Arpa Services which is the Tcp/ip stuff.  Network Services is HP proprietary.
HP's naming conventions of these products (NS vs NS-Arpa) is not very helpful
in clarifying that there are TWO different protocols running.

Hope this is helpful
Steve Mestad, stevem%pserv@src.honeywell.com

dfc@hpindda.HP.COM (Don Coolidge) (08/18/89)

>I've a copy of RFC 1034 (Domain Concepts and Facilities) and have
>checked therein before starting this mild flame...

>As you may notice from my address, we have a three part domain
>name following the node name. It seems to work well here and
>we see no reason to change.  However, HP seems to think that
>one should only have a two level domain name, e.g. HP.COM.  When
>I issue the command "nodename anon.cs.colostate.edu", I get a
>message "nodename: invalid node name syntax".

I'm afraid you're confusing two similar things. The machine name
entered via the nodename command has nothing to do with domain names.
It's the HP-only nodename, required only for HP-proprietary Network
Services (NFT and RFA). It is limited to three sections (two levels)
by HP specification. Unless you make use of NFT and/or RFA, you have 
no need to ever execute "nodename". The nodename is never used by
any non-HP-proprietary code. It has nothing to do with RFC 1034.
HP nodenames were in use long before the RFC came out.

I'm not sure how you enter the BIND domain name for your machine. Full
support for BIND first appeared with the 6.5 release on the HP9000 s300.
In any case, the "nodename" command isn't how you do it.

By the way, you have a three-part domain name following your "hostname"
(by HP-UX terminology), not your "nodename".

>Back to RFC1034... check section 3.1 last paragraph and the example in
>3.5 for mention of a three part domain and size limits. Also note
>RFC1035 sections 2.3.4 and 3.1 for more on size limits.

>I suspect that a really aggresive competitor could disqualify HP as a
>bidder for failing to meet this spec.  Has this happened?  (Remember
>DEC getting bumped from the DOD procurement a year or so ago, based on
>their failure to provide the "standard" Unix?)

I don't see any aggressive competitors disqualifying us for meeting
our own proprietary spec ;-). Rather, I'd imagine they're more dismayed
to see us implementing more industry standards with time, instead of
sticking to proprietary solutions.

>Yet more flaming is based on the hostname length restrictions.
>While this isn't as well founded as the other part, it's still
>irritating to come upon limits (of eight characters), seemingly
>based on tradition.

It's based on System V compliance. HP-UX is based in part on both
System V and BSD flavors of un*x. System V specified the eight-character
limit. In later releases (6.5 onward), that limit's changed. 

- Don Coolidge

The above comments represent my own views as a friend to the Net, and
should not necessarily be taken as HP's official pronouncements. When
in doubt, call your Support Center. (Of course, they'll agree with me ;-) )

raj@hpindwa.HP.COM (Richard Jones) (08/18/89)

You are prossibly running into restrictions imposed by the NS software
and not by the 'BSDesque' software.  NS (that wonderful beast created
for the HP 3000 MPE systems) uses 8 character pieces in a three
piece name
              node.domain.organisation

please keep in mind that while this looks like a 'domain' name and might
resolve if used as a domain name, it ain't a domain name ;-)

I'll let one of the UX gurus expand on and correct this as my knowledge
is much better in the MPE world...

rick jones

dfc@hpindda.HP.COM (Don Coolidge) (08/19/89)

Slight correction : full BIND support doesn't appear in 6.5. It's on its
way, though. Ask your HP rep to find out when.

(Sorry for the previous posting error - it's been a long week...)

Don Coolidge

harkin@hpindda.HP.COM (Art Harkin) (08/19/89)

A minor correction to Don Coolidge's comment about the availability
of BIND (Domain Name Services) on hp9000 computers, it will be 
available in the soon upcoming 7.0 HP-UX release and is not currently
available for 6.5 (as previously stated).

A last bit of clarification, the 3 part (level) 'nodename' has 
absolutely nothing to do with Domain Name Services, in fact it is
not used by ARPA/Berkeley services and NFS, it is ONLY used by HP's
proprietary services (NFT and RFA). Unfortuately, this has been
a point of confusion for several customers. 

Art Harkin
Hewlett Packard

"No, we did not implement a lobotomized version of BIND"

fritz@hpfclp.SDE.HP.COM (Gary Fritz) (08/22/89)

Just as proof, Randy, check out my address.  We've been running with a
"three-level" node name for some time -- at least since we upgraded to
HP-UX 6.5, and probably earlier.

Gary Fritz

diamant@hpfclp.SDE.HP.COM (John Diamant) (08/24/89)

From: Gary Fritz (fritz@hpfclp.SDE.HP.COM)
> Just as proof, Randy, check out my address.  We've been running with a
> "three-level" node name for some time -- at least since we upgraded to
> HP-UX 6.5, and probably earlier.

As Art Harkin also pointed out, 7.0 is the first release with full
domain (BIND) support.  Gary didn't realize that we were running pre-release
code on hpfclp.sde.hp.com in order to get domain support.  Sorry for the 
confusion.


John Diamant
Software Engineering Systems Division
Hewlett-Packard Co.		Internet: diamant@hpfclp.sde.hp.com
Fort Collins, CO		    UUCP: {hplabs,hpfcla}!hpfclp!diamant