In the past I have always used Kate on a remote Linux server without any problems. I now have a new configuration:
- new Linux server running Slackware 14.1 with KDE 4.10.5
- Windows 7 laptop running Xming and Putty

When I start Kate on the server from my laptop it is very very slow. top on the Linux server shows sshd working hard. In Kate I have to wait 10 second for a menu to scroll down. It is useless to try to use the keyboard as characters take seconds to display! Essentially it is useless!

I have tried to run other programs including KDE specific programs and they work fine (some with higher latency but still acceptable). The connection to sever is on LAN at 100 Mbit/s. For example emacs works very fast. Typing in xterm also is very fast. dolphin and kwrite work fine too. Konqueror has more latency but it is usable. Firefox menus are very fast to navigate.

I have also started KDE locally on the server and tried Kate. It is rather slow even locally but very much usable. I have remove all display effects, etc. The server has limited graphics (it's a VME interface board) so I expect local graphics to be slow. What is very difficult to understand for me is why particularly Kate is so slow when I access it remotely from my laptop. My laptop is a Lenovo W530 a little bit older than one year so I don't expect resource problems.

Could anybody point me to what can be the problem and how to solve it? Thanks!

kate -graphicssystem native

In Qt5 this is gone though, so maybe you should look for a remote desktop client?

Thanks a lot scummos! On my installation "kate -graphicssystem native" resulted in a very large speed increase. Kate is now usable remotely. Could you briefly explain why is much faster with this flag? Cheers!

X across network works by redirecting the X protocol calls to the remote server, and having that execute it. So for example if you want to draw a rectangle, you'd send a call to X which says "draw a rectangle". Redirecting that over the network is fast. That's what happens when you use -graphicssystem native. This is a technique pretty much from the 90's though -- today, clients want to do drawing themselves usually.

With the default graphicssystem (raster) a pixmap is drawn by the application itself (so, on the server!) and then a call is made to display that pixmap on the screen. Using that over the network causes the pixmap to be transferred to the client, which is of course much more effort than just transferring the command.

Take all that with a grain of salt, I'm not an X person

Greetings!