Re: ra-privacy: my responses to comments

[Brian E Carpenter <[e-mail]>]

On 02/08/2013 01:26, Scott Brim wrote:
> On 08/01/13 14:31, Keith Moore allegedly wrote:
>> There are many people (in IETF and elsewhere) who believe that
>> applications should never use IP addresses directly or in referrals to
>> other applications.   This is often cited as if it were some
>> architectural principle - in fact just last night, I actually had an AD
>> state that to me as if it were a principle.   I happen to disagree
>> emphatically with that supposed principle, for many reasons, but I won't
>> list those reasons here.   
> 
> You may have to if you want to press your case.

Some of Keith's arguments on this point were captured in
draft-carpenter-behave-referral-object-01, but that draft also
contains the statement:

"  In some cases, this problem may be readily solved by passing a Fully
   Qualified Domain Name (FQDN) instead of an IP address.  Indeed, that
   is an architecturally preferred solution [RFC1958]. "

That's indeed what RFC 1958 says.

>> For the moment it only matters that there is
>> a widely held belief that all applications should only use names to
>> refer to hosts or application endpoints.   From that point-of-view, all
>> hosts/nodes need to have names, so (by this definition) all hosts/nodes
>> need to have public addresses.   
> 
> All sources of Internet public services need to have DNS names, but
> that's it.  Other than that, "names" are only needed in higher layer
> communications, and can be handled there.  For example, your laptop
> doesn't need a name to open communication with a SIP server, but once it
> does it can use one or more SIP-level identifiers for its end of the
> SIP-level communication.

It can, but applications can get into trouble for a whole lot of reasons
discussed in the above draft and its successors (draft-carpenter-grobj-reqts
and draft-carpenter-referral-ps).

   Brian