Re: The death John McCarthy - LISP, HIP & GSE

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

 



Hi Jari,

Thanks for clarifying your position.  When I wrote "you are both" the
"you" applied only to Luigi and anyone else to appears to be happy with
referring to the LISP protocol as something like "the Locator -
Identifier Separation Protocol".

The "both" was not meant to include you - it applied to the two points
which followed:

  assuming that LISP is not only a Loc/ID Separation protocol,

  but "*the* Loc/ID Separation protocol".

You wrote:

> Speaking just for myself, I have not assumed that LISP is the only
> identifier - locator separation protocol. Apologies if something that I
> wrote mistakenly gave that impression. 

> I didn't get this impression.

> As you know, there are several
> identifier - locator separation architectures, with very different
> designs and characteristics. It would be inappropriate to claim that
> there is just one, or that one particular definition is the only
> possible interpretation. 

OK.  In the thread which I think is best titled for this discussion:
"Re: LISP is not a Loc-ID Separation protocol", I suggested that GSE,
HIP and now ILNP and lesser known RRG proposals definitely are Loc-ID
Separation protocols.

If no-one disputes this, then I think those who support the LISP
protocol retaining its current name should argue persuasively either
that the LISP protocol operates on similar enough principles to GSE, HIP
and ILNP, or that the definition of "Loc-ID Separation" should be
expanded to include the principles by which the LISP protocol works.

I think the first option is unsupportable and that expanding the
definition so far as to apply to the LISP protocol would render it
largely meaningless.  If such a redefinition was generally agreed to, I
would continue to argue against it, since it would also apply to Ivip.
The most fundamental principles of LISP, Ivip and IRON are similar - and
I argue that they are entirely different to, contrary to, and
incompatible with the principles on which GSE, HIP and ILNP operate.


> Indeed, the very definition of "identity" could range from something
> that looks very much just like another address (LISP, Mobile IP,
> HBA-based SHIM6) to new identifier concepts (HIP, CGAs) or even to
> DNS names.

I think DNS could be regarded as a Locator - Identifier Separation
protocol.  However, it doesn't seem helpful to group it with GSE, HIP or
ILNP, since these three apply to "every host-to-host communication".
DNS is not required for host-to-host communications, but it can be used
to choose which host IP address(es) to communicate with.

With the LISP protocol the hosts deal with IP addresses which embody
both the ID and Loc functions, just as IPv4 and IPv6 works today.  I
think the same is true at the application level and perhaps the stack
level, depending on which form of Mobile IP is used.

My initial impression of the IPv6-only Hash Based Addressing RFC-5535 is
that these 128 bit values function identically to today's IP addresses
as far as the host stack and the routing system.  They do not embody any
purely ID functionality - so there must be some separate arrangement,
implemented by applications and presumably suitable support in the
stack, to work with a multi-homed host (so with multiple HBA addresses)
on the basis of an Identifier which is presumably not an IP address.
The most obvious way of doing this would be to use the host's DNS name
as its Identifier.  However, for this to be used for all communication,
I think the responding host, receiving a packet from some novel IP
address, would need to do some kind of reverse lookup in order to obtain
a DNS name, and then another lookup of that to get
locators, from which to choose from when sending its response.
Otherwise, the recipient host must respond directly to the original
packet's source IP address, without knowing for sure that this can be
used as a locator for the host whose DNS or other form of Identifier is
presumably in the application-level part of the original packet.

So my initial impression of HBA is that it could be used to support a
Locator - Identifier Separation protocol, with suitable application and
stack changes, if the Identifier namespace was to be other than the DNS
namespace, or if responding hosts were also expected to find all the
Locators of a multi-homed originating host before replying.

As far as I know, everyone agrees that HIP is a Locator - Identifier
Separation protocol.  I understand that HBA is a particular way of using
CGA.  As far as I know, CGA and/or HBA are examples of the sorts of
mechanisms you could use in building a complete Locator - Identifier
protocol, but which do not in themselves constitute a complete such
protocol with the aim as GSE, HIP and ILNP: to apply to "every
host-to-host communication".

I recall a difficulty with ILNP and I guess GSE and HIP not being able
to use their own multihoming protocol for the necessary mapping lookups
- so by "every host-to-host communication" I mean all application and
stack-level communications other than these lookups which are intrinsic
to the Loc-ID Separation protocol itself.

  - Robin
_______________________________________________
Ietf mailing list
Ietf@xxxxxxxx
https://www.ietf.org/mailman/listinfo/ietf


[Index of Archives]     [IETF Annoucements]     [IETF]     [IP Storage]     [Yosemite News]     [Linux SCTP]     [Linux Newbies]     [Fedora Users]