I guess that you can only have a control that can be implemented in HTML and rendered on a browser. The only way I could think of is something like what you are suggesting.

I am considering a table, spanning rows and columns to simulate frames, and then very narrow, empty, solid-colored spanning rows and columns that work as pseudo-splitters. I could listen for mouse events for the pseudo-splitters in JavaScript and change the cursor and dynamically move splitter rows and columns and resize table cells on the browser to simulate splitter behavior.

I need the user to be able to manually resize areas in the browser so that they can allocate their screen real estate for the various pseudo-frames. Real frames make state management too difficult when frames have tightly coupled data and behavior.

I will probably handcode the JavaScript in the page and then, maybe someday, I could code the control that actually would automatically generate the JavaScript and tables for the programmer who wanted to use it.