Re: [v6ops] discussion of transition technologies

[Lee Howard <[e-mail]>]

On 1/21/18, 2:28 PM, "Fred Baker" <[e-mail]> wrote:

>Restated. Close?
>
>> On Jan 19, 2018, at 4:22 PM, Fred Baker <[e-mail]>
>>wrote:
>> 
>> At least part of this commentary wound up in
>>https://tools.ietf.org/html/rfc6180
>>     Guidelines for Using IPv6 Transition Mechanisms during IPv6
>>     Deployment. J. Arkko, F. Baker. May 2011. (Format: TXT=49679 bytes)
>>     (Status: INFORMATIONAL) (DOI: 10.17487/RFC6180)
>> 
>> I think Jari's view in that was that we needed to rein in the plethora
>>of transition technologies, and "if one has to translate, can we please
>>do so above the IP layer?" I added SIIT/NAT64, because I think there is
>>market relevance including several deployments of various kinds; any
>>mention of MAP-T or 464XLAT is SIIT/NAT64. But the basic recommendation
>>of RFC 6180 was:
>>  - first choice, deploy native IPv6 - for scenarios in which IPv6
>>islands are connected across IPv4 space, use dslite (a tunneling design).
>>  - for scenarios in which IPv4 islands are connected across IPv6 space,
>>use dslite (a tunneling design).

You would argue that 464xlat, MAP-E, and MAP-T are simply implementations
of SIIT/NAT64 (which you do, below), but it seems to me that they are also
used in these scenarios. In some cases the “island” is but a single
application on a mobile phone.

For large networks, like access ISPs, mobile networks, and maybe campus
networks, the advice might be finer-grained:
- for scenarios in which IPv4 islands (e.g., households, a mobile phone,
or other stub network), consider:
 — If you need home gateway support right now, use DS-Lite
 — If you are using CGN anyway (such as mobile networks), consider
464xlat, which allows native IPv6 end to end when possible, native IPv6
end-to-translation-edge when IPv4 is needed to reach the other end, and
CLAT on the device or IPv4 edge when the local application or device
requires it.
 — If you can pressure or wait for home gateway vendors for MAP support
and would prefer stateless BRs (probably more scalable/cheaper than
stateful DS-Lite AFTR), use MAP. (One more decision branch between MAP-T
and MAP-E, but they’re pretty close).
 

>  - for scenarios in which IPv6 islands are connected across IPv4 space,
>use 6rd (a tunneling design).
>>  - for scenarios in which IPv6 systems have to talk with IPv4 systems,
>>translate. Please consider doing so above the IP layer.
>> 
>> I would argue that 464XLAT, MAP-E, and MAP-T are "services in which an
>>ISP might use SIIT/NAT64 in its network", and are therefore not
>>fundamental transition technologies as much as ISP services built using
>>them.
>> 
>> I think I might also argue that the market has more or less followed
>>that advice. Your spreadsheet seems to suggest that.
>
>The interesting thing is that 6rd, which is a way of appearing to have an
>IPv6 network without actually having one, is not what one might call
>"prevalent". It has in fact been used for *transition*, in places like
>Free - which used to connect IPv6 customers using 6rd and (I understand)
>has recently announced native IPv6 deployment. The places I know that
>have used it used it for a while and then have gone native.
>
>Would you agree with that?

I would; that is my perception. MHO is that 6rd has had its day, and while
I don’t think it needs to be deprecated, I haven’t heard any scenarios in
the past several years where it solves an actual problem.


Lee