Re: Automatic selection of transition mechanism?

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

On 2008-05-08 21:58, Nathan Ward wrote:
> On 26/03/2008, at 1:41 AM, <teemu.savolainen@nokia.com>
> <teemu.savolainen@nokia.com> wrote:
> 
>> Hi,
>>
>> As we now have a set of mechanisms to transport IPv6 over IPv4
>> standardized, and as it seems we will probably end up having a set of
>> mechanisms to access IPv4 destinations over IPv6-only access (MNAT-PT,
>> SNAT, NATv4v6v4…), has there been discussion already how a host that
>> implements multiple transition technologies should efficiently and
>> automatically discover which mechanisms to use in access networks it
>> happens to attach (and move between)?
>>
>> Is the current state of the art just: trial & error, provisioning by
>> non-IETF means, or as defined by other SDOs for their systems?
>>
> There hasn't been a reply to this yet, so I'll fill the void.
> 
> I haven't kept up to date on the new stuff that people are talking
> about, SHANTI, (M)NAT-PT and so on.
> 
> Windows does a fairly reasonable job, by doing roughly:
> - Look for ISATAP.<domain>. If it exists use ISATAP.
> - If we have a non-RFC1918 address use 6to4.
> - Last resort use Teredo.
> 
> That seems to be a reasonable approach. The one big bug is if the
> non-RFC1918 address is filtered, NATed, etc. 6to4 will generally fail to
> operate.

Right, which suggests a need to encourage reachability checking so that
black holes are detected.

I think it's too soon to think about operational pragmas for IPv4-via-IPv6
until the various solutions settle down.

   Brian