Hi,

while writing my dissertation related to the topic, I couldn't help spotting few issues from draft-perkins-dmm-matrix-03 in section 3.7, "HIP/LISP".

I realize that you're tight on space especially in Table 1 but I think HIP and LISP focus on different problems, and it would be useful to separate them into different sections. HIP focuses on mobility and security where as LISP tries to improve routing scalability. Also, the standardized HIP is deployed on end-hosts where as LISP is deployed on edge routers.

HIP is "backhaul" friendly as it's deployed to the end-host. When an application on a HIP-capable host communicates with HIP-incapable host, it just uses regular IP addresses instead of HITs. Thus, no proxy is needed and the following statement is incorrect in section 3.7:

  4.  backhaul friendly: HIP/LISP both use routing address for packet
       routing; there is no centralized anchor point in the data plane.
       But for communication to other hosts which are not located in the
       HIP/LISP domain, a gateway function is needed and the data
       traffic is constrained to travel through the gateway.

As an author of RFC6317, I'd like to state that it is not widely used yet (if at all) and it is completely optional. The legacy-compatible way of using HIP as specified RFC5338 is the prevalent way of using HIP. So I would say that the following statement about HIP is incorrect (and the "may" statement was further incorrectly applied in table 1):

   5.  app friendly: LISP does not require application modification.
       HIP may require application modification [RFC 6317].

There is something wrong with XML as points 7 and 8 are somehow convoluted:

   7.  Local routing: For communication within the HIP/LISP domain, HIP/
       LISP can support local routing since the routing is based on
       routing prefix instead of host indentification and there is no
       cent

   8.  ralized anchor point.

Regarding to point 7, HIP can route local traffic locally as it is an end-to-end protocol.

Finally, the table 1 in section "Matrix Comparing Existing Approaches for DMM" should be adjusted according to my corrections:

backhaul friendly: should be Y for HIP (and probably N for LISP)

app friendly: should be Y for HIP and Y for LISP

local routing: should be Y for HIP (and Y for LISP, at least according to the explaining text)

Thanks!
_______________________________________________
dmm mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/dmm

Reply via email to