John,

> The purpose of network addresses is to locate an entity in the graph of 
> the network.  Network addresses are intended to facilitate forwarding.  
> For forwarding, only the structure of the graph is important.  (Notice I 
> did not say "topology."  A topology refers to the invariant properties of 
> a mapping among sets. A graph is not a topology. A set of graphs might be.

> Yes, I know the field abuses the term all the time. Another failure of 
> education.)

I do understand that the intial purpose of layer-3 is to route data through
topological map, but that was 30 years ago, the IP concept is even older, I
believe it's about time to look into what has been done so far, and to let
some changes take place, and to give more space to new applications and
services to be introduces by providing the proper infrastructure.


> Physical location has nothing to do with the operation of the network 
> layer. It is all about forwarding packets within the graph formed by the 
> layer.  When one says that network addresses are location dependent (and 
> they are) it is with respect to the graph of the network.  This is also 
> why addresses must be assigned *by* the network.  Only the network knows 
> where in the graph the nodes are.

Once you provide location-based services, physical and network locations
will be strongly binded together, otherwise network will not know where to
forward these packets and topological location alone will be useless.


> I believe you were referring to Google here.  The wonderful thing about 
> software is that it is possible for not very good programmers to do dumb 
> things that will work for awhile.  As we have seen in this field since its

> inception, it is dangerous to confuse economic success with good science.

> "Well, it works!" is merely an excuse, not an argument for good science.

Good science should be practical enough to be applied in real life, business
and services invest technology to make peoples life easier and more
prodective .. I may not agree with you in refferring to google and others as
"not very good programmers to do dumb things that will work for awhile".


Thanks,
Ammar




-----Original Message-----
From: John Day [mailto:[email protected]] 
Sent: Monday, November 19, 2012 2:39 AM
To: Eitan Adler; Ammar Salih
Cc: [email protected]; [email protected]
Subject: Re: Adding GPS location to IPv6 header

It has been interesting to watch this discussion, but Eitan and others have
it correct.

The purpose of network addresses is to locate an entity in the graph of the
network.  Network addresses are intended to facilitate forwarding.  For
forwarding, only the structure of the graph is important.  (Notice I did not
say "topology."  A topology refers to the invariant properties of a mapping
among sets. A graph is not a topology. A set of graphs might be. Yes, I know
the field abuses the term all the time. Another failure of education.)

Physical location has nothing to do with the operation of the network layer.
It is all about forwarding packets within the graph formed by the layer.
When one says that network addresses are location dependent (and they are)
it is with respect to the graph of the network.  This is also why addresses
must be assigned *by* the network.  Only the network knows where in the
graph the nodes are. 
(Yes, this means that so-called MAC addresses are not addresses. They are
serial numbers.  Neither were IP addresses pre-CIDR.)

All identifiers in computing are addresses and locate something.  As Saltzer
says, to resolve a name is to locate an object in a given context.  Things
like a so-called flat address is really just an address  used outside its
context.  For example, MAC addresses locate the manufacturer and however
else the manufacturer used the space.

This is all pretty basic stuff.

At 18:22 -0500 2012/11/11, Eitan Adler wrote:
>On 11 November 2012 17:15, Ammar Salih <[email protected]> wrote:
>  >They use IP address instead, and it's not always about http 
>applications, how about VoIP applications, now you need another 
>mechanism?

I believe you were referring to Google here.  The wonderful thing about
software is that it is possible for not very good programmers to do dumb
things that will work for awhile.  As we have seen in this field since its
inception, it is dangerous to confuse economic success with good science.

"Well, it works!" is merely an excuse, not an argument for good science.

Take care,
John Day

>
>Nothing stops application layer protocols from sharing one mechanism
>
>>  .. how about detecting your preferred language for layer-3 routing?
>
>Why does language matter ever matter in the network layer?

It has nothing do with physical location either.  So, if you would put that
in why not something else.

>
>>>as there are many countries with more than one popular language,  not 
>>>mentioning that many ip registrations does not even reflect the 
>>>traffic originating country.
>
>Exactly. You can't infer language from location. Some locations have 
>multiple valid languages.  This isn't a valid use case.
>
>>  I've explained this in previous parts of the document, mainly 
>>because Layer-3 devices won't be able to recognize the feature, and 
>>also to unify the location implementations at different layers.
>
>Why do layer 3 devices need this information?  So far you have 
>mentioned exactly one use case that *might* be useful.
>
>>  It doesn't have to be always .. at least now you partially agree :)
>
>I didn't say that. I said that this is the first time I you have 
>mentioned anything that *could* (not necc. should) be acting at layer 
>3.
>
>>  Users currently have absolutely *NO* control over IP<->location 
>>mapping, it's totally how your IP owner has
>>  registered the IP subnet, what I am suggesting is that your local 
>>ISP *can* tag the city location "if it's required", unless you want to 
>>share your exact location or set the location to all zeros, in this 
>>case you are asking the ISP not to tag your location, but in this case 
>>you give up all location based services.
>
>Unless the ISP decides to ignore your "request" and provide the 
>information anyways. Also, what if your actual location differs from 
>the IP address range your request is coming from. This can easily occur 
>from tunnels or mobile networks.
>
>>>  Response: It does not have to be in every IPv6 header, only when 
>>> there  is location update, also the host should have the option of 
>>> not to  send location updates.
>
>How does the IP layer know when the *application layer* needs this 
>information?
>
>>*BUT* if it's required by the government/or any other
>>>organization or third party in the future for the sake
>>of >protecting the copyright laws then the feature will be >available 
>>to support that as well.
>
>How would my ISP know my GPS coordinates?  Once again please think 
>about this in the context of tunnels and mobile networks.
>
>>  Google.fr example is confusing many people, which I will modify, 
>>policy based routing has much more than routing tcp:80 traffic.
>
>Why should the policy differ based on GPS coordinates?
>
>
>--
>Eitan Adler
>--------------------------------------------------------------------
>IETF IPv6 working group mailing list
>[email protected]
>Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6
>--------------------------------------------------------------------

--------------------------------------------------------------------
IETF IPv6 working group mailing list
[email protected]
Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------

Reply via email to