hi Yiu,

sorry but i just found i missed a technical issue you left in this message.

2012/4/9 Lee, Yiu <[email protected]>

>
>    on the other hand, 4rd-u makes a tight coupling between the address
> format
>    and its own header mapping mechanism. this makes an operator unable to
> have
>    different choices of transition mechanism as long as it chooses 4rd-u.
> if
>    you concern the choice of IPv6 transition mechanism, i do recommend
> MAP, as
>    either encapsulation or translation is operatable with MAP address/port
>    mapping without difficulty.
>
> [YL] Sorry for my lack of knowledge. Please explain how MAP isn't coupling
> address format and header mapping?
>

tight coupling here, i mean, that 4rd-u header mapping mechanism sometimes
depends on features of the address format, like the CNP and the V-octet.
for example, it removes the L4 checksum adjustment, this makes the whole
system MUST following the addressing with CNP.

for the MAP, things are different. boxes are implemented with existing
header processing modules but only need to add MAP address mapping module
into them. once the operator deploys MAP following its addressing plan
paradigm (we are still working on a better guidance for this in the MAP
deployment draft), the operator can choose either MAP-T or MAP-E for the
header mapping without the need of changing their addressing plan. if the
operator find translation is suitable for this part after a certain period
of practice on encapsulation, it can switch the mode by informing the CE to
change the mode of working (surely sometimes such an action of informing is
not easy in practice for, e.g., the home CE subscribers but it is feasible
for enterprise networking). another example of de-coupling is translation
approach: RFC6052 defines a stateless address mapping while RFC6145 defines
the header processing which is usable for both stateless and stateful
solutions, both 1:1 case for non-renumbering transition and residual
deployment of MAP.

implementation does also benefit from the de-coupling. the RFC6145
translation module can directly work with address mapping module of MAP or
RFC6052. the MAP address module can work with header processing module of
RFC2473 (for MAP-E) or translation module of RFC6145 (for MAP-T).

it doesn't mean the tight coupling doesn't work. like the Burton snowboard,
it works fantastically with its own system of bindings, not interchangable
for other brands ;-) however, in most of time, we really think
interchangability, supported by loose coupling, is preferred.

cheers,
maoke
_______________________________________________
Softwires mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/softwires

Reply via email to