Re: [DNSOP] I-D Action: draft-bellis-dnsop-session-signal-02.txt

[Bob Harold <[e-mail]>]

On Thu, Jul 28, 2016 at 1:28 PM, Ray Bellis <[e-mail]> wrote:

> Notwithstanding the ongoing Call for Adoption, we've just posted a
> further update to this draft which addresses some of the comments
> already received, and also removes the explicit "Start Session" TLV in
> favour of just using the "Idle Timeout" TLV to request a "long-lived
> session".
>
> Ray
>
>
 Just my thoughts - what seemed 'normal' 10 years ago seems terribly slow
now, so I hate to specify what is "short" or "long" times in a long-lived
document.

2.  Terminology - "session" says:
"The connection between client and server is persistent and relatively
long-lived (i.e. minutes or hours, rather than seconds)."

I am thinking that even 'seconds' could be long-lived.  Would it be better
to say that a session could be "long-lived" in terms of "time or number of
packets or round trips" ?

4.2.1. Idle Timeout
"It is not required that the Idle Timeout TLV be used in every session.
While many Session Signaling operations (such as DNS Push Notifications
[I-D.ietf-dnssd-push]) will be used in conjunction with a long-lived
connection, this is not required, and in some cases the default 30-second
timeout may be perfectly appropriate."

I would be interested in the list discussing whether there should be a
default time or not.  Would it make sense to leave the default timeout up
to the server, and always send an Idle Timeout request if a 'long-lived'
session is desired?  I might be wrong, but am curious what others think.

-- 
Bob Harold