Since ZK 6, in addition to PollingServerPush and CometServerPush ZK provides a third Server Push implementation - Asynchronous Processing CometServerPush. As their name suggest, they implement the Client-Polling, Comet (aka., long-polling) and Servlet 3 Comet (Servlet 3 Asynchronous Processing-based Comet) server pushes. Client-Polling is available in all editions, Comet Server Push is available in ZK PE and EE, while ZK EE supports Servlet 3 Comet Push.
The default implementation depends on which ZK edition you use:
|PE||Comet Server Push|
|EE||Servlet 3 Comet|
You can also configure ZK to use the one you prefer, or even use a custom server push.
Choose an Implementation
Client-polling is based on a timer that peeks the server continuously to see if any data to be pushed to the client, while Comet establishes a permanent connection for instant push. Client-polling will introduce more traffic due to the continuous peeks, but Comet will consume the network connections that a server allows.
You could configure a particular ZK page to use a particular implementation by the use of DesktopCtrl.enableServerPush(ServerPush). For example,
((DesktopCtrl)desktop).enableServerPush( new org.zkoss.zk.ui.impl.PollingServerPush(2000,5000,-1));
If you would like to change the default server push for the whole application, you could use the the server-push-class element as follows.
<device-config> <device-type>ajax</device-type> <server-push-class>org.zkoss.zk.ui.impl.PollingServerPush</server-push-class> </device-config>
where you could specify any implementation that implements ServerPush.
The client-polling server push is implemented with an implicit timer at the client. The interval of the timer depends on the loading of the server. For example, the interval becomes longer if the time to get a response has become longer. PollingServerPush uses a timer to peek if the server has any data to push back. The period between two peeks is determined by a few factors.
The minimal delay to send the second polling request (unit: milliseconds).
The maximal delay to send the second polling request (unit: milliseconds).
- The delay factor. The real delay is the processing time that multiplies the delay factor. For example, if the last request took 1 second to process, then the client polling will be delayed for
1 x factorseconds. Default: 5.
- The larger the factor is, the longer delay it tends to be.
It could be configured in
WEB-INF/zk.xml by use of the preference element as follows.
In additions, you could specify them in the constructor: PollingServerPush.PollingServerPush(int, int, int). For example,
((DesktopCtrl)desktop).enableServerPush( new org.zkoss.zk.ui.impl.PollingServerPush(2000,10000,3));
Comet Server Push Config
The comet server push is implemented with a pre-established and 'virtual' permanent connection. It is like sending a taxi to the server, and waiting in the server until there is data to send back. Meanwhile, the client-polling server push is like sending a taxi periodically to the server, and leaving immediately if no data is available.
CometServerPush comes with its own configuration parameters such as retry delay, retry count, and ajax timeout.
Configure these parameters by using the preference element in your
<preference> <name>CometServerPush.retry.delay</name> <value>3000</value><!-- 3 seconds delay between each retry--> </preference> <preference> <name>CometServerPush.retry.count</name> <value>3</value><!-- 3 tries for each request --> </preference> <preference> <name>CometServerPush.ajax.timeout</name> <value>180000</value><!-- 3 minutes --> </preference>
The minimum time delay to send the next comet request retry (unit: milliseconds).
The maximum retry counts if a comet request failed, -1 means keep retry forever (unit: count).
The amount of time a comet request will wait for a server response until it aborts. (unit: milliseconds).
Default: varies, depending on browsers
The configuration of the errors is handled by the client-reload element, specified in
WEB-INF/zk.xml. The markup below demonstrates an example of catching an error of the server push:
<error-reload> <device-type>ajax</device-type> <connection-type>server-push</connection-type> <error-code>410</error-code> <reload-uri>/login.zul</reload-uri> </error-reload>
where the connection-type element specifies through which channel the requests are sent. By default it is the AU channel in which Ajax requests are sent by widgets running at the client. If you would like to specify the error page for server-push then connection-type must be set to
When using ServerPush the client engine will send "dummy"-requests (without extra payload) to the /zkau servlet to pull queued server side updates.
Those requests look similar to this:
In the case of PollingServerPush there will be one of these requests per configured interval, sometimes causing a firewall to give a false alert.
In the case of CometServerPush, after a long polling comet request returned, there will be one "dummy" request sent to retrieve server update.
Since ZK 8.5.0, the WebSocket based Server Push WebSocketServerPush is provided. When WebSocket connection is enabled, this will be used as default, other configuration specifying which Server Push to use will be ignored.
|6.0.0||Feb 2012||The CometServerPush is available in ZK PE, while ZK EE supports Servlet 3 Asynchronous Processing-based Comet.|