[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: addition of TLV to locator ID or locator ID set




El 03/10/2005, a las 15:58, Paul Jakma escribió:

On Mon, 3 Oct 2005, marcelo bagnulo braun wrote:

i am not sure what you mean here, nut just for stating things precisely, HBA do have globally assigned prefixes

Uh, ah. I didn't realise that. Who assigns them?

HBA are constructed using regular PA prefixes assigned by the ISPs (LIRs) that are serving the multihomed site. Just regular PA prefixes. The HBA essentially defines the iid

 And surely the point of HBA is that they can be locally assigned?

Well, the iid is locally generated (as any IPv6 address) but the prefix is asigned by the LIR as any PA prefix

(Or do you mean HBA's stand a very good chance of being globally /unique/?).


no, HBA are administratively unique (as opposed to statistically unique)

I had, in my mind, been considering that there would be a mechanism to allow the prefix set -> ULID mapping to change

there is, the CGA part of the HBA/CGA hybrid addresses.

That is why there are three flavors of addresses:
- HBA that support a static set of addresses that cannot be changed during the lifetime of an established communication (with the cost of a hash per address) - CGA/HBA that support a stable prefix set and additional addresses that can be added during the lifetime of the communications (the cost of the stable addresses is a hash operation while the cost of a new address (out of the initial set) is a public key operation) - CGA addresses, any addresses can be used with the cost of a public key operation

I'd have to ask: Why use CGA?

i think that Jari and Erik have clearly explained this point


AFAICT, CGA was created specifically because relying on IPSec for address assignment has bootstrap issues. Hence CGA and SeND.

For modes relying on anonymously introduced public-key cryptography, we won't have bootstrap issues here, so why not use standard IPSec? Is it just to avoid the size of an AH header on each packet?

How will you communicate the CGA parameters? Is shim to reinvent portions of IKE in order to communicate public keys? Why not just use IKE? IPSec is well-understood, widely implemented and deployed. Why give ourselves the task of reinventing aspects of it?

I can see some of the attractions of HBA, but using CGA beyond the local-link is stretching it..

(That's not to say a host that was configured with a key for CGA couldn't use the same public key for IPSec..)

However, you should note that the public key used is included in the hash that results in the interface identifier of the addresses

Ok.

fairly big cost in functionality. (Eg, what exactly would shim6 using invariant HBAs offer over application layer remapping support?

i am not sure i understand what you mean here... could you expand?

For me, the prime benefit of multihoming is not so much having multiple links but having number portability and the ability to easily transfer between providers[1]. Eg, shim'ing but having only one provider would be something I'd be interested in if it gave me portability. It's been a big reason for multihoming for two organisations I've worked with.


i agree that this is a great benefit of current IPv4 PI address based solution, but this is clearly not the goal of the shim6 wg


Without portability, you're left only with solving failure events amongst multihomed links.

I would rephrase it a bit more general, like: the goal of the shim is to preserve established communications through outages but yes, portability in the sense of PI addresses are not part of the scope of the shim work

 And I don't think you need a shim layer for this, personally.

here we disagree (but we have already discuss this point)

regards, marcelo



regards,
--
Paul Jakma	paul@clubi.ie	paul@jakma.org	Key ID: 64A2FF6A
Fortune:
travel, n.:
	Something that makes you feel like you're getting somewhere.