Re: Loopback interface terminology issue

[Toerless Eckert <[e-mail]>]

If i followed BrianC's note correctly, then the loopback interface
dates back to BSD Unix stack earlier than routers: The problem
was how to allow the use of TCP/IP if you do not even have any
external interface up and running. YOu didn't want to create/choose
yet another socket type for that case. Would have made apps very complex.

Are historic BSD 4.2 vs. 4.3 sources freely available now for some history check ? ;-))

On Fri, Oct 20, 2017 at 05:33:13PM +0100, tom p. wrote:
> ----- Original Message -----
> From: "Michael Richardson" <[e-mail]>
> To: <[e-mail]>
> Sent: Wednesday, October 18, 2017 5:52 PM
> 
> > Brian E Carpenter <[e-mail]> wrote:
> >   brian> Well yes; indeed this topic is touched on in many RFCs but
> nowhere is
> >   brian> it defined as part of the basic architecture, which is my
> main
> >   brian> point. Using a concept that has no principal definition is
> generally
> >   brian> a source of confusion.
> >
> >   joel> We tended not to define software interfaces for things which
> are not
> >   joel> required for interoperability. How you describe an address
> which is
> >   joel> not bound to a physical interface is entirely irrelevant
> outside the
> >   joel> scope of your own operating system.
> >
> >   brian> That's not our experience in trying to be precise in saying
> what we
> >   brian> mean in draft-ietf-anima-autonomic-control-plane.
> >
> > to restate Brian's point slightly differently:
> >
> > The issue is not to tell people how to implement things in their
> software,
> > but rather to make it clear that we need a standard hook on which to
> hang our
> > addresses.
> >
> > We'd rather not use the term "loopback interface" at all here if we
> had
> > another name defined in an RFC somewhere.
> >
> > I think that there is some significant text that might need to be
> written
> > when defining this hook when in a strong-host model.
> >
> > But, let me also ask a different question: are there router operating
> systems
> > where there is a way to hang an address on something which is not a
> > virtual/loopback interface?  If so, what do they call it?
> 
> When I first came to IP, I was surprised to find that L3 addresses were
> assigned to interfaces, as opposed to nodes, the latter having been the
> case with other networks I had used previously.  I have always seen the
> creation of a virtual interface as a quirk which overcomes this absence
> of a node address, or node identifier in the IP namespace.
> 
> I first encountered the term 'loopback interface' in Cisco
> documentation, which talks of the ability to configure a 'virtual
> interface, referred to as a loopback interface'.  So for me, 'loopback
> interface' has always been the Cisco term for a virtual interface but
> why it should be loopback, as opposed to virtual, I have never found
> out.
> 
> Whatever the name is, it does provide an identifier for the node and
> one, at least in some systems, that is always up and so provides a
> stable identifiier for protocols that need one, such as a router-id in
> BGP; and you can have more than one such identifier, separating
> functions, perhaps by priority, to prevent DoS attacks..
> 
> Tom Petch
> 
> > If there are differences in software interfaces in some place that we
> would
> > be respecting by abstracting this requirement?
> > After 35+ years of router operating systems, has anyone actually
> innovated in this way?
> >
> > --
> > Michael Richardson <[e-mail]>, Sandelman Software Works
> >  -= IPv6 IoT consulting =-
> 
> 
> --------------------------------------------------------------------
> IETF IPv6 working group mailing list
> [e-mail]
> Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6
> --------------------------------------------------------------------

-- 
---
[e-mail]