Re: [SWMP] Re: faster field messages

> 
> So each side gets to declare its own mapping and the other is obligated to 
> keep track of that.

What are the limits, like a maximum number of mappings, or a
maximum guaranteed lifetime of mappings? Unlimited mappings
that each peer needs to keep track of open the gate for easy
DoS attacks. Negotiated limits with each peer would not allow
a server to use the same mappings for each client.
One approach would be that a peer advertises how many mappings
it is going to use, so that the partner can decide on whether
to accept that number or cancel the connection. Another is to
define a maximum number in the protocol, like 255 or 256. That
may be too restrictive, but allowing two bytes and more than
65000 mappings will trigger memory problems, in particular on
a server.

>From the paper (I've read it by now) I perceived it rather
as an ad hoc technique used to shrink the size of a group of
messages in an UDP packet. In that case, the lifetime of a
mapping could be restricted to the packet. But that doesn't
map to TCP connections, or your discussion of defining the
mappings on a separate channel.
In the extreme case of ad hoc, you'd have only one mapping
which is defined at the beginning of a sequence of messages
referring to the same node, and overwritten by the next
definition. Actually, that's not even a mapping anymore,
but an implicit argument.

cheers,
  Roland

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