dhcpcd-discuss

On Thu, Mar 16, 2017 at 4:25 PM, Roy Marples <roy@xxxxxxxxxxxx> wrote:

> On 16/03/17 16:57, Shahid Mahmood wrote:
> > On Thu, Mar 16, 2017 at 10:38 AM, Roy Marples <roy@xxxxxxxxxxxx
> > <mailto:roy@xxxxxxxxxxxx>> wrote:
> >
> >     On 16/03/2017 13:48, Shahid Mahmood wrote:
> >     >     Why would you want to configure it anyway?
> >     >
> >     > We are seeing a situation where upon a RENEW, the server sends a
> >     > different IP address than requested.
> >     >
> >     >
> >     >     You'll get the new config
> >     >     eventually (when T1 expires or the carrier goes down/up),
> reconfigure
> >     >     accept just gets you it faster when pushed from the server.
> >     >     Because the protocol requires a nonce, it's pretty secure.
> >
> >     Can you explain more why this is a problem for you given the above?
> >     With DHCPv6, you'll just get the new address faster and the existing
> >     address just won't get refreshed and the kernel will remove it when
> it
> >     expires.
> >
> > Basically the application requires persistent IP sessions at run time.
> > An IP address change would disrupt those sessions, leading to a longer
> > recovery path.
>
> Ah, now we get to the real issue.
> So dhcpcd has the "leasetime" option, which is only used for IPv4
> addresses currently.
>
> We could re-use this to send an empty address with the valid and
> preferred lifetimes set to the "leasetime" option - so you effectively
> request the address for that amount of time. The specification allows an
> all-ones lifetime to represent infinity.
>
> Would that satisfy your need?
>

I think we are good with things as they are for now. This is not a critical
scenario to justify any change.
Thanks for sharing the idea! Really appreciate your input.


> Roy
>

-shahid