James,

Let me see if I have the functional nature of RISP correct (haven't worked with 
it before):

1) The devices that RISP hosts have permanent non-routable IP address space. 
RFC 3103 says RFC 1918 space....I'm assuming this would by ULA space in IPv6.

2) The RISP Gateway has a pool of availble PA addresses that it can assign.

3) The RISP host negotiates a short lease (4 or 5 seconds) for one of these PA 
addresses from it's RISP gateway and binds it.

4) The RISP host then uses that PA address to communicate directly for the 
duration of the (4 or 5 second) lease.

Is that essentialy correct?

If it is, this leads me to a few questions...

 - Either the the RISP host must communicate with a communication partner that 
has a staticaly assigned IP to initiate a communication session or the RISP 
hosts must have some out-of-band method to exchanging thier leased IP info.... 
else how would one RISP host ever initiate a communication session with 
another? Furthermore, how would the recieving RISP host know to obtain a PA 
lease without some out of band method for communicating that it was about to 
recieve an incoming communication?

 - If there is some out-of-band regsitration of RISP hosts then that 
effectively defeats the purposes of "obscuring" the devices IP address since 
another RISP host can query the out-of-band service to obtain the IP address 
for each RISP host whenever it wants to do so. Although this might not be 
horrible depending upon the kind of security required to query the out-of-band 
service.

 - What happens when the RISP host recieves INBOUND communication on it's 
leased PA from a device which is not it's current communication partner? Under 
dynamic NAT this would go nowhere since there isn't an entry in the NAT flow 
table that corresponds to that session. Am I left with a single point of 
failure (my FW filter rules) for stopping these...or does the RISP gateway 
functionaly stop it?

 In effect, the only way I could see this practicaly working would be if the 
RISP effectively permanently held open a lease...and either that lease was long 
term....or if it did change every few seconds then there would have to be some 
sort of automatic registration scheme with some sort of out-of-band registry.

In either case, it wouldn't be doing what I want...since I don't want my hosts 
accessible/addressable unless THEY initiate communications. If my hosts are 
registering with some sort of out of band service then I've just introduced 
another method of profiling my devices...unless that service has a really 
air-tight method of establishing trust relationships. If the leases are 
effectively semi-permanent then it doesn't sound much more obscure then 
DHCP....and would seam to have alot of the same overhead and drawbacks.

It doesn't sound like a terrible solution...but I still don't see why I would 
prefer it over NAT. It sounds like it has many (though not all) of the same 
drawbacks as NAT, may miss some of the functionality NAT provides.... and 
doesn't sound like it has any less overhead or complications that implimenting 
ALG's in NAT.

If NAT provides me exactly the functionality I want now....doesn't break any 
protocols/applications I want to use. The applications/protocols it breaks I 
ACTUALY WANT broken.... why would I want to switch to RISP instead?

But thank you for the information on it....it was interesting regardless.






















Christopher Engel


> -----Original Message-----
> From: james woodyatt [mailto:[email protected]]
> Sent: Friday, April 30, 2010 4:16 PM
> To: Chris Engel
> Cc: NAT66 HappyFunBall
> Subject: Re: [nat66] Terminology: Definition for "IPv6 Realm"?
>
>
> On Apr 30, 2010, at 13:26, Chris Engel wrote:
> >
> > You are not going to achieve that level of "obscurity" without some
> > form of address translation....
>
> That's not true.  It can also be done with encapsulation.
> RFC 3103 is an existence proof.
>
> > and any solution that you do provide to achieve that obscurity will
> > have much of the same side effects that todays NAT does.
>
> Also not true.  As RFC 3103 shows.
>
>
> --
>
> james woodyatt <[email protected]>
> member of technical staff, communications engineering
>
>
>
_______________________________________________
nat66 mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/nat66

Reply via email to