Re: Issue with retry_count in socket.c and fix for the same

On 05/15/2014 02:25 PM, Heikki Linnakangas wrote:
> I wonder why we need any timeout or retry counters, ever. Frankly, I'm
> inclined to just rip all that out. A timeout would be a useful feature
> for many applications, to detect a broken connection more quickly, but
> the way it's currently implemented is pretty much useless for that. For
> starters, it would have to be configurable, because many applications
> would not want to have a timeout at all. A built-in timeout of 60
> seconds or so, which I think you'd get with your patch, would be too
> short for OLAP kind of queries that can run for hours before returning
> any results.

Ok, I removed the broken logic, for the sake of cleaning up the code. It
still won't give the kind of a timeout that you hoped for, but at least
it's less bogus now.

- Heikki