draft-mickles-v6ops-isp-cases-05.txt

[Jun-ichiro itojun Hagino <itojun@iijlab.net>]

>3.3.1  IGP
(snip)
>   Case 2: Existing ISIS IPv4 network
>
>   If the ISP chooses to build IPv6 capabilities using OSPFv3, then 
>   considerations of existing hardware and memory constraints must be 
>   made since OSPFv3 places additional load on network gear.  This 
>   IGP will operate in separate memory space and will need to be configured 
>   separately from any existing ISIS implementation.  
>
>   If the ISP chooses to build IPv6 capabilities using ISIS, then the 
>   amount of hardware resources are not as taxing since IPv6 is 
>   integrated within ISIS.  The protocol does not need to be configured 
>   separately.

	Even if you chose to run IS-IS for both IPv4 and IPv6, you still need
	to maintain IPv6 routing table with IS-IS, which will have memory
	constraints to routers.  We could share some information (like router
	adjacency) among IPv4 IS-IS and IPv6 IS-IS, but how significant it is
	to share such information compared to run IPv4 IS-IS + OSPFv3?  The
	above text seems to me very IS-IS centric view.  Is there any
	quantitative measurement on this topic?

>3.3.4  Multicast
>
>   PIM-SM is the generally accepted solution for deploying multicast.
>
>   For IPv6, this is a hard problem.

	Even for IPv4, deployment of PIM-SM is a hard problem.

	This section does not talk about multicast routing across different
	ASes - what will we use for this?

editorial:
	why some of the section names in section 5 are all uppercase?