On 2 Nov 2016, at 1:09, Jerry Lundström wrote:
Hi Paul,
On 11/01/16 16:48, Paul Hoffman wrote:
- It feels like combining both class and type into ClassType might be
over-optimization. Since Class will almost always be IN, why not just
have this as its own object member?
I was also looking at this and there are some values which are very
common so you could add to the specification "if not specified then X
is
Y" assumptions.
Arrrgh. I think I now understand that Class/Type tables (Section 7.10)
are a way to say "there are a very limited number of combinations of
Class and Types, so let's combine the two into single entries for
pointing to". If I have that right, then making ClassType a list would
save two bytes per entry in the table.
Current:
ClassType = {
type => uint,
class => uint,
}
type = 0
class = 1
Proposed:
ClassType = [ type, class ]
I cannot imagine a case where a query or a response doesn't have both a
class and a type.
--Paul Hoffman, who totally admits that this totally dives into the
format and not the picture of whether it is useful and to whom...
_______________________________________________
DNSOP mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/dnsop
