Re: draft-v6ops-van-beijnum-mnat-pt-00.txt

[Brian E Carpenter <brian.e.carpenter@gmail.com>]

On 2008-02-14 10:56, Rémi Després wrote:
> Iljitsch van Beijnum a écrit :
>> On 13 feb 2008, at 14:58, Rémi Després wrote:

...
>>> One of the items would be a better understanding of why mapped
>>> addresses have been avoided in your proposal.
>>
>> Because some people REALLY don't want to see them on the wire
> Technical reasons should be known for the debate to progress.
> 
>> because it could be confusing to have the same address block used for
>> two different purposes depending on the presence of native IPv4
>> connetivity.
> Unclear to me wo far.

It's simply an operational nightmare. Imagine trying to explain
a problem with such an address to a help desk.

> 
>>> Another item would be where checksum adjustment best fits.
>>
>> Let me know if you have any ideas about that.
> 
> When a  address is replaced by another, adjusting the checksum is simple.
> Using ad hoc values in prefixes in order to keep the checksum change to
> zero seems to me more complex than needed.

Nevertheless this forces all packets to go through extra packet
inspection and possible computation cost. I think this has performance
implications for the MNAT-PT box.

> It may prevent other uses of prefix fields, that may be more useful.

Well, putting semantics in address bits is not such a good idea
in general.

    Brian