> Yes, I see that, but what about the protocol *insists* that it must be
 > so?

Nothing.

 > I'm fishing around for a fundamental reason why some parameters are
 > special and others are not.  I don't see that reason, and everything
 > you're citing is revolving around the implementation we have.

It's not revolving around our DHCP implementation, it's revolving around
our DHCP architecture.  The architectural decision was to have the client
manage only the essentials and leave the rest to dhcpinfo.  While one
could argue about what is essential, at this point that ship has sailed.
I don't know why you find it uncomfortable to base other decisions on this
architecture.

I agree administrators may be surprised by this, and if we (actually, the
guys we contracted to do the original work in 2.6) had to do it over again
maybe it would be done differently.  But I'm not trying to redesign that
architecture, I'm just using it as a guide to keep the overall behavior of
the DHCP client consistent.  To me, it's consistent to have the DHCP
client watch over the parameters that it is documented to configure.

-- 
meem

Reply via email to