Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!zaphod.mps.ohio-state.edu!pacific.mps.ohio-state.edu!linac,att!ucbvax!UMDD.BITNET!BRUCE
From: BRUCE@UMDD.BITNET (Bruce Crabill)
Newsgroups: comp.protocols.tcp-ip
Subject: Re: What is INADDR_LOOPBACK for in sockets?
Message-ID: <9103181807.AA02191@ucbvax.Berkeley.EDU>
Date: 18 Mar 91 13:42:10 GMT
Sender: daemon@ucbvax.BERKELEY.EDU
Organization: The Internet
Lines: 29

>1) On my IBM TCP/IP for VM, "own-address" goes through the network, as you say.
>2) But trying FTP 127.0.0.1 gets me talking to the first non such router on the
>   path to the Internet.
>3) 14.0.0.0 is used instead to stay local.
>4) and, of course, ftp 14.0.0.0 from a host whose Internet path I am on gets
>   to my system.
>
>So, it looks like the specific address is implementation dependent.
>A question is: how can a procedure (script or whatever) wishing to use the
>loopback address be made portable when there is no "localhost" to refer to
>in the dns?

In this respect, the IBM TCP/IP (FAL) is broken (maybe it has been fixed
in version 2).  The "Assigned Numbers" RFC (RFC-1010) states:

   (g)   {127, <any>}

      Internal host loopback address.  Should never appear outside
      a host.

I.e., any class A network address that starts with a 127 in the first
octet is a loopback address.  This has been brought to IBM's attention
in the past and the response is that it is this way for historical
purposes and compatibility to previous versions.  The fact that it will
actually place a 127.x.x.x number on the network is just plain broken.
If it isn't fixed in version 2 (which is just now getting out to people),
I will try to make a SHARE requirement of it at the next SHARE meeting.

                                       Bruce