2013/2/5 Ole Troan <[email protected]>

> > But there has been a firm requirement in the WG that, even for shared
> addresses, one of the CE could have the well-known ports.
> > The WKPs-authorized option has been added to 4rd to satisfy this
> requirement.
> > I don't think this requirement has disappeared.
>
> does anyone in the working group have an opinion on the well-known ports
> and shared IPv4 addresses?
> the options we have are:
>
> a) if well-known ports are required then a full IPv4 address must be
> assigned
> b) allow that one user gets the well-known addresses,
>     but then the port block must at least be > 1024 ports.
>

(a) is comfortable. i do think that, when MAP is used for non-shared IPv4
addresses, either public or private, well-known ports are surely available
with the exclusively used IPv4 addresses to enable server-side
applications. such a demand exists. the idea is as the same as (a).

(b) is acceptable with the condition that the port set size > 1024.
well-known port usage for the shared IPv4 address is somehow tricky as it,
philosophically, doesn't fit some considerations when we design the MAP
architecture. e.g., well-known ports are well-known while the A+P framework
incline to assign non-contiguous port numbers for a host for some reasons
like privacy/security. on the other hand, the port set size > 1024
condition may restrict the operation so that PSID length, offset, ea-bit
length cannot be selected arbitrarily according to the regular MAP
deployment logic.

- maoke


> do we know that end users only need ports 0-1023 or 0-4095?
> or are the ports the require to be opened inbound across the range?
>
> I'm leaning towards a.
>
> cheers,
> Ole
> _______________________________________________
> Softwires mailing list
> [email protected]
> https://www.ietf.org/mailman/listinfo/softwires
>
_______________________________________________
Softwires mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/softwires

Reply via email to