Matthew Dovey wrote:
> No - the original model was that you had to do something to get a new TTL
> (e.g. ask for the next ten records in which case the server would give you a
> result set name/TTL) We did have a debate against absolute vs relative time.
> I favoured the former but the general opinion was that we weren't doing
> mission critical timing here so the latter was easier and sufficient.
Absolute vs. relative is not the current question, it's ttl vs. idleTime. (A ttl
could be absolute or relative; an idle time would always be relative.) Yes I
think we decided on relative vs. absolute, but the fact that we were even
discussing absolute makes me think we had ttl in mind rather than an idle time.
But I don't think we ever made this distinction.
I think Ralph would prefer an idle time to a ttl. I don't have a preference.
Does anyone else? (If not we'll go with idle time.)
An idle time and ttl would differ as follows: if it's an idle time the server
might supply it initially, say 10 minutes, and each time the result set is
accessed the 10 minutes automatically starts again without the server having to
update it (though the server could supply a new idle time to supercede the old
one). If it's a ttl, after 10 minutes the result set expires, no matter how much
activity, unless the server updates the ttl.
And I suppose we have to decide both for a result set and for a session, and one
could be an idle time and the other a ttl.
--Ray
|