Re: Fwd: A comment about MAST

["Eugene M. Kim" <gene@nttmcl.com>]

Dave,

On Fri, Sep 12, 2003 at 02:19:03PM -0700, Dave Crocker wrote:
> 
> NAT boxes do address translation and MAST does address translation. In a very
> theoretical sense there is, therefore, some similarity between them. However
> MAST is not a "network" address translation service and it really has nothing
> to do with the controversy surrounding NATs.
> 
> So I am not understanding criticism of MAST that is essentially a continuation
> of the criticism about NATs.  They are separate and I think the debates should
> be kept separate.

Agreed.

> 
> As for a "MAST API", none is needed.
> 
> MAST hides the multiple addresses from transport and above.  That is one of
> its features.

I see.  Looks like I misunderstood MAST in that sense. =)

> 
> I believe that any application protocol that uses direct IP addresses has a
> problem with any multi-homing/mobility solution.
> 
> Also, as soon as the application needs to benefit from knowing about -- and
> giving guidance about -- the availability of multiple addresses, then it needs
> an be nice to enhanced API. That is true for any multi-addressing solution,
> not just MAST.

Also agreed.

A random idea: How about making MAST support these two operating mode?

1. The original scenario, where the transport layer such as TCP does not
know how to handle multiple endpoint addresses nor requests MAST to
notify about address changes.  MAST operates as originally proposed in
this case.

2. The new scenario, where the transport layer such as TCP knows how to
handle multiple endpoint addresses and requests MAST to notify about
address changes.  MAST then propagates such changes up to the transport
layer, which, in turn, may decide to propagate such changes up to the
applications via a suitable API.  What I meant by `the MAST API' was,
in fact, this.

Thanks,
Eugene