Re: noid and applications (generic requirements from applications)

[Brian E Carpenter <brc@zurich.ibm.com>]

Patrik,

In NOID the transport layer and above aren't aware that the locator
is rewritten. So I don't see that part of your concern.

More fundamentally, you say

> Today, the idea is that the IP address as well as the FQDN is globally
> unique

Unfortunately that certainly isn't true of addresses today and we still have to
fight hard for it to be true of IPv6 addresses tomorrow. It's a dangerous
assumption. However, there's no reason FQDNs should be non-unique. 

    Brian

Patrik Fältström wrote:
> 
> On 2003-11-13, at 14.48, Tony Li wrote:
> 
> > That is going to be exceedingly difficult.  As long as we
> > are using IP addresses in applications and we are hiding
> > the actual reachability of the address from the application,
> > we are guaranteed to be allowing an application to pass along
> > an unreachable address.
> >
> > Note that this is a result of the layering violation of
> > passing around an IP address in the application.  Asking us
> > to warp the world to support architectural mistakes strikes
> > me as suboptimal.
> 
> The reason I ask is because I want to know what we will get at the end
> of the day with the noid proposal.
> 
> Today, the idea is that the IP address as well as the FQDN is globally
> unique and have the properties described.
> 
> |    (A=B) and (B=C) gives (A=C) where '=' is "can communicate with"
> |
> |  and
> |
> |    (A->B) gives (B->A) where '->' is "can send data to"
> 
> What I understood from the noid draft is that what we today call IP
> addresses (what is above L2, used in the arp requests etc) can be
> rewritten and because of this doesn't have the properties above.
> 
> My question was whether the new identifier will have the properties
> above, so an application instead of doing layering violations on
> FQDN+IP should do the same kind of violations on FQDN+Identifier?
> 
> Else it will be extremely hard to "fix" protocols like ftp and
> everything which controls RTP streams.
> 
> I think the discussion whether those protocols are designed in a
> correct or wrong way will not help, as you and myself possibly have the
> same view -- even though there are rumors there are differences in how
> hot it is in hell.
> 
> IF the above properties are valid for at least one of the identifier
> "layers" in noid, fixing applications and protocols we have today will
> be much easier.
> 
>     paf