status (Thursday soon)

[Alistair Woodman]

So I don't think that Alain is "deluded" about the fact that resetting the port range is not a "violent" action. Operationally I don't think it would happen more than once every 6 months or so, and then at 2am in the morning during maintenance. SPs would understand the issue.

I think we are making too much about the "user disruption" issue. It is no worse that an SP bouncing the link, which also happens every few months...

Sent from my iPad

On Mar 22, 2012, at 12:51, Francis Dupont <fdupont at isc.org> wrote:

>> <snip>
>> => 01 is compatible with 02, the problem is in the other way.
>> => I implemented the 01 and setup the demo in the intersection of 01 and 02
>> (by accident: it was just the simplest, good for a demo but not for the real
>> world but this is another debate :-).
>> 
>> So your sdnat code implements 01 but is configured to behave as 02 ?
>> (because 01 is a superset of 02?)
> 
> => yes
> 
>> <snip>
>> => this won't be in the demo and IMHO it won't work (i.e., deeply broken)
>> 
>> Is the issue more "bad protocol / systems design" or is the issue with
>> "using ICMP within the Linux system and the ability to re-adjust ports on
>> the fly" ?
> 
> => the first unfortunately.
> 
>> I.e. are your objections design or coding specific ?
> 
> => The design. I don't say there is no coding issue but this is not
> dependent on the way a port range change is announced, just from the
> idea of port range change and the illusion it can be smooth...
> 
> Regards
> 
> Francis Dupont <fdupont at isc.org>