Just to tighten up the parameters a little...
Portability is not an issue but the display (screen is not always relevant with some of the proposals) should be person sized IE that you it can sit in front of you while you work.
I'll note several of you came up with a version of the split flap display which would be reasonable.
A few tidbits that might help stimulate creativity.
1/3 of all pen nibs in the world during the 19th century were made in a 5 sq mile area around Birmingham. These nibs were high precision sheet metal components with tolerances in the thousandths of an inch range, mfgd by a multiple stage pressing process.
OTOH devices that rely on controlled levels of friction seem more doubtful
At around that time Reuters where transmitting dispatches around the world using microfilm messages attached to pigeon tails, so photo-etching would appear to be a viable technology for this era after all.
Rather more doubtful is a thing called a Pockels Read Out Memory (PROM). This is based on Bismuth Silicon Oxide. It's possible to read and write it with light. The crystals are fairly small so some kind of projection system (like a microfiche reader) would probably be needed. A focused light light spot on the crystal at the right color could do selective erasure
When I first thought of this challenge I pictured the back end being something like a the hardware in side a 1970's photo typesetter, where the various fonts were stored on microfilm cards and the font size varied with an optical system.
Those who pointed out something like this was described in "The Difference Engine" should remember that was for a cinema sized projection TV type system, no good for interactive I/O. Standard I/O was by punched cards and printer listings, not really what I was thinking of.
The etch-a-sketch ideas might also be viable as a graphics terminal for CAD work (sort of), although I'd guess more as an out put device with most of the "design" being by processing a command language, rather than tracking a mouse or light pen.
The real challenge is the temporary and partially erasable image storage problem.