Machine names on the net ...

[David Herron -- a slipped disk]

In article <25902 at cup.portal.com> thad at cup.portal.com (Thad P Floryan) writes:
>forrie at suntau.UUCP (Forrie Aldrich) in <33 at suntau.UUCP> writes:
>	I am confused about names of computers... I have been told that you can
>	only have a maximum of 6 characters for a machine name... 
...

I'm afraid that there's enough mis-information here that I must step in and
correct what I can..  Generally, Thad, I enjoy reading your postings but
this time you're a bit off the track.

>...  I've used PDP-10, Tenex, DEC-20, Foonly (all related) systems
>since their inception (the PDP-10 back around 1965), and one interesting
>aspect of it was that, for "efficiency", filenames were stored in a single
>36-bit machine word allowing up to 6 characters in a filename.  My conjecture
>is THIS is the genesis of the 6-character machine name.            ^^^^^^^^^^

Good thing you labeled this as conjecture..  I fail to see how a
strangeness on a non-Unix OS could affect naming of machines in a
network of what amounts to being Unix BBS's.  I know TOPS-10 quite well
(incidently, they fit 6 characters into 36 bits by using a restricted
character set that fit into 6 bits/character.  tricks like this are
easy on the PDP-6/10/20 architecture since it allows for varying byte
sizes)

>The Usenet, as such, started circa 1980 (if memory serves), and was required
>to be in conformance with Arpa protocols (e.g. email formats, and so forth,
>based on what's known as RFC 822).

Yes, 1980.  But, RFC-822 wasn't written until a couple years later.
Also the earlier versions of Usenet -- remember that we're on the B
version of Usenet right now, earlier version was A -- used a very
different header format.  And again within the B version there was
a number of changes in format over time.  Earlier versions of B had
(conscious?) incompatibilities with RFC-822, for instance Article-ID:
instead of Message-ID:.

The compatibility with RFC-822 came in around the time of B v2.11
and RFC-950.

>For whatever reason(s), all the early protocols permitted up to 6-char names,
>and much extant uucp software still abides those restrictions.  More
>recently, HDB (aka BNU) (circa 1984) uucp software permits up to 8-character
>names.  I "believe" it's possible names can be longer (your "snorklewacker",
>for instance :-) but they must be UNIQUE within the first 6 (or 8)
>characters.

There's a couple places where a limit can come from:

-- Usenet software & what it can put into the sys file, and Path: lines.
   I wouldn't be surprised if early versions of Usenet software had a
   limit on this name.  I'd also be surprised if there were *still* a
   limit.  There isn't any intrinsic limit since the format is "name!name!..."

   Note that the name for Usenet purposes is (can be) different from your
   UUCP or domain names.  Especially since a Path: line looks so much like
   a UUCP route-address that people insist on believing that it contains
   useful information for routing e-mail.  It is best if you use the same 
   name everywhere...  but the software doesn't require this.

-- UUCP software has had a variety of limits on host names.  Usually in
   the range of 6-8 characters.  There's no really good reason for this
   other than early development was done on PDP-11's and the software from
   that era shows all sorts of strange limits supposedly to make the software
   *fit* in the first place.

   Funny, this 3b1 I'm on right now seems rather small (memory wise) -- 
   especially when I think about porting X to it.

-- Arpanet names.  I don't believe there was ever a limit, except for
   what people were willing to type.  In fact, there were numerous
   machine names which were rather long when I started in on the nets
   in '84 ..  a whole bunch of 'em at CMU with names like

	cmu-cs-foo-baz.arpa

   Then the naming pattern for gateway machines is to name the connected
   networks and tack "gw" on the end.  Like uky-sura-gw.uky.edu.

   There was some interaction between arpanet hosts and Usenet hosts in the
   early days of Usenet.  In fact, my first lesson on the uselessness of
   Path: lines as e-mail addresses was while I was trying to use Path:
   lines to get UUCP routing information.  I learned real quick that the
   machine named "vgr" wouldn't pass any UUCP mail through it, despite what
   it said in the Path: line.(vgr.arpa is now == vgr.brl.mil)
-- 
<- David Herron; an MMDF guy                              <david at ms.uky.edu>
<- ska: David le casse\*'      {rutgers,uunet}!ukma!david, david at UKMA.BITNET
<- 
<- New official address:  attmail!sparsdev!dsh at attunix.att.com