Re: [Ila] Identifier size

For LISP, you can assign an EID to the loopback interface, all 128-bits. And then the interface addresses that are either statically conifgured or learned by SLAAC are 128-bit RLOCs.

You can assign multiple EIDs to the loopback interface, be them crypto-EIDs or not, or a combination of either.

If ILA (or ILNP) useds 64-bit identifiers, those can be regsitered to the LISP mapping system and return 128-bit RLOCs. Or for that matter, return any size you want. To be used by how any data-plane wants to use the addresses.

Dino

> On Jan 31, 2018, at 9:12 AM, Tom Herbert <tom@herbertland.com> wrote:
> 
> 
> 
> On Wed, Jan 31, 2018 at 8:27 AM, Behcet Sarikaya <sarikaya2012@gmail.com> wrote:
> Hi Tom, all,
> 
> I changed this tread to identifier size issue.
> 
> What is the motivation for crypto-graphic identifiers?  Is the idea to give each device a master identifier and then it can use the a crypto graphic function to independently create its own unique identifiers for use in communications. That would be good for address per connection and 80 bits might be doable in ILA.
> 
> Saleem pointed out that:
> ILNPv6 will not work with more than 64 bits in the NID, and that is consistent
> with RFC8200/STD86 (which refers to RFC4291, for the use of a 64 bit ID).
> 
> 
> So my question is the identifier in identifier - locator separation equal to the interface id in RFC 8200?
> 
> No, it's not. This is where one of the problems with identifier locator address split arises. SLAAC performs /64 address assignments. This is assigning a  subnet to a device with the expectation that IIDs in the subnet (lower 64 bits) are assigned by the device receiving the assignment,  Many mobile providers use SLAAC to assign /64 to UEs. This is in contrast to using DHCPv6 to get singleton addresses. The IID space is used by the UE for assigning addresses to downstream devices (like in tethering) as well randomizing address for local binding as a means to mitigate address scanning attacks (address scanning was used in WannaCry attack). In this sort of address assignment it's the upper sixty-four bits that identify the mobile device, the identifier for identifier/locator split would be derived from the upper sixty-four bits.
> 
> Sixty-four bits isn't enough to encode both a locator and identifier, but I think a level of indirection will work. This is my description of that:
> 
> A device may be assigned a /64 address via SLAAC as is common in many provider networks. In this scenario, the low order sixty-four bits contains IIDs arbitrarily assigned by devices for its purposes; so these bits cannot be used as an identifier in ILA. The alternative to support /64 prefix assignment is to encode an identifier in the upper sixty-four bits. Since only a subset of bits are available, a level of indirection is used so that  when ILA transformed the upper sixty four bits contains both a locator and an index into a locator (ILA-N) specific table. The entry in the table provides the original sixty-four bit prefix so that ILA to SIR transformation can be done.
> 
> If yes, then what happens if the UE has more than one interfaces?
> 
> This makes it the uniqueness of the IID and the identifier is the same problem?
> 
> In ILA, identifiers need to be unique with an ILA domain. Normally, this will mean it is unique with one SIR prefix. That is analogous to an IID being unique within a subnet.
>   
> Tom
> 
> _______________________________________________
> ila mailing list
> ila@ietf.org
> https://www.ietf.org/mailman/listinfo/ila