Hi all, the other day I worked a bit on the app_server... I have a pending patch that changes the way updates are sent to the client. The problem with the current implementation is this: When the client code invalidates multiple views or just different parts of one view while the message queue is blocked, then the server will send _UPDATE_ for the first invalidation, while the remaining invalidations will be combined in the server. So there are two update sessions triggered. On R5, only one is triggered (I think). To make it work like on R5, several things could be done: 1.) The server is notified by the client that the client is currently "busy", ie the server should wait with sending update requests until the client has told the server that it is now ready to receive them. :-\ 2.) The server sends update requests for all invalidations, they are combined in the client when the messages are processed. :-| 3.) The server doesn't put any information in the _UPDATE_ message at all and simply tells the client that there would be an update pending for it. When the client receives the notification, it asks the server what stuff needs updating. The server could have combined new update information in the meantime. I implemented 3.) and it improves drawing speed quite a bit. WonderBrush is much more responsive now. The problem is that some drawing defects can be observed (small areas not being redrawn sometimes). Until I have more time to investigate this bug, should I still commit the patch anyways, since it makes redraws quite a bit faster? Best regards, -Stephan