[dhcwg] Lifetime draft: refresh time should never be more than IRT_DEFAULT

In reading the lifetime draft, I am generally _very_ happy with it.   
However, there is one point that I think is a mistake - allowing the 
time to be set to an unreasonably large value.   In considering this, I 
think it's useful to look at other configuration parameters that IP 
hosts frequently acquire, such as the IP address of a server.   When 
acquiring the IP address of a server, we go to the DNS and do a lookup. 
   It would be extremely unusual for the result that we get back to have 
a TTL that is more than a week, and I think pretty commonly the TTL 
would be more like a day, or even an hour.

DHCPv6 Information Request transactions are like DNS transactions.   
They do not require a database write, so they are relatively 
low-impact.   They can be implemented in routers.   Basically, there 
are ways to make them really, really cheap.   So optimizing the 
protocol with the intention of making them happen really infrequently 
seems like a mistake to me.

I think that the maximum refresh time a client should ever accept is a 
single day.   I can imagine someone arguing for a larger time, such as 
a week, but I think that generally the only reason you'd ever get a 
time that big would be because someone was trying to spoof your client.

Aside from this comment, I think that it's past time to advance this 
draft - it looks very good.

_______________________________________________
dhcwg mailing list
dhcwg@ietf.org
https://www1.ietf.org/mailman/listinfo/dhcwg