Re: Architectural argument for RAs [Re: [69ATTENDEES] DHCP]

[james woodyatt <jhw@apple.com>]

On Sep 21, 2007, at 13:20, Iljitsch van Beijnum wrote:
> There are three questions:
>
> 1. Is there a reason why it would be helpful to run a network with  
> routers but without RAs?
> 2. Would 1. be compelling enough to live with the loss of fate  
> sharing it implies?

This assumes that fate sharing is viewed by all participants as a  
feature, not a bug.  I'm not sure I should believe this is a valid  
assumption.

> 3. Would it be possible to set up hosts with default configuration  
> that works both in currently deployed IPv6 networks and networks  
> configured as per 1.?

For some definition of "works"... remember, the next thing the DHCP  
people will want after they get default gateway and prefix length is  
to demand that DHCP servers can be authoritative for this  
information, not the routers, i.e. the DHCP servers will say to  
nodes, "turn off router solicitation and stop processing  
advertisements for the duration of your lease."

You might find that DNAv6 is required to kick nodes into giving up  
their DHCP lease and retrying router solicitation before they will  
"work" properly, i.e. just like with IPv4.


--
james woodyatt <jhw@apple.com>
communications software engineering
1 infinite loop; ms: 306-2TC; tel: x49423