Single color or tiled pattern wastes no system memory for background buffers. Color value or bitmap (actual bit map) was written directly to VRAM each time some region needed to be filled with background, and VRAM was the only component that needed to be big enough to hold the screen buffer. A single byte value or a couple of bytes was enough to define the whole background in the program.
Later, small tiled bitmaps (pix maps) were used for the same reasons to fill the screen. They were compact enough to fit somewhere in discardable free memory, and also to read from the disk with a single operation, which meant there was no benefit in keeping them in memory backed by swap file.
Some modern distributions use programmatic gradient generation for desktop background, though I am not sure whether it still works directly, or simply creates a temporary full screen image in memory, which is no different from using a regular photo.
Later, small tiled bitmaps (pix maps) were used for the same reasons to fill the screen. They were compact enough to fit somewhere in discardable free memory, and also to read from the disk with a single operation, which meant there was no benefit in keeping them in memory backed by swap file.
Some modern distributions use programmatic gradient generation for desktop background, though I am not sure whether it still works directly, or simply creates a temporary full screen image in memory, which is no different from using a regular photo.