Update TCP keepalive docs

doc/keepalive.md

@@ -16,3 +16,5 @@ backend (or network equipment) will interfere with these connections, it's bette
 ## TCP Keepalives
 
 The keepalive mechanism above is ideal for long-standing idle connections, but it cannot be used during query processing. With some PostgreSQL-like data warehouse products such as [Amazon Redshift](http://docs.aws.amazon.com/redshift/latest/mgmt/welcome.html) or [Greenplum](http://greenplum.org/), it is not uncommon for a single SQL statement to take a long time to execute, and during that time it is not possible to send `SELECT NULL`. For these cases you may want to turn on *TCP keepalive*, which is quite different from the application-level keepalive described above. To better understand the different kinds of keepalives, see [this blog post](http://blog.stephencleary.com/2009/05/detection-of-half-open-dropped.html). As that article explains, TCP keepalive depends on networking stack support and might not always work, but it is your only option during query processing.
+
+On Linux, you turn keepalives simply by specifying `Tcp Keepalive=true` in your connection string. The default system-wide settings will be used (for interval, count...) - it is currently impossible to specify these at the application level. On Windows, you can also specify `Tcp Keepalive Time` and `Tcp Keepalive Interval` to tweak these settings.