Under the Hood: Building CatchSpace with Zero Memory Overhead on Windows

When we set out to build CatchSpace, we had one non-negotiable rule: A desktop utility designed to sit idle in the background must never behave like a web browser.

In 2026, the software industry suffers from an epidemic of bloated micro-utilities. Everyone has experienced downloading a trivial productivity tool—a clipboard history manager, a color picker, or a snippet shelf—only to discover that it consumes 400MB of RAM, spawns seven background processes, and requires a dedicated GPU rendering context just to display a 200-pixel floating rectangle.

CatchSpace was designed to solve a fundamental operating system annoyance: moving files between overlapping windows on Windows 10 and 11 without endlessly resizing or alt-tabbing. But to justify its existence as a permanent resident of your desktop, its footprint had to be imperceptible. Here is how we engineered CatchSpace from the ground up to operate under 38 megabytes of working memory.

1. The Native Dilemma: Electron vs. Pure Win32

The path of least resistance for cross-platform utility development is Electron or Chromium Embedded Framework (CEF). You write some HTML and CSS, throw in Node.js for file I/O, and ship an installer. The cost, however, is disastrous for utility software:

  • A baseline memory footprint of 150MB–300MB even when completely idle.
  • Slow cold-start times exceeding 1.5 seconds.
  • Inefficient OLE drag-and-drop interop that stutters when dragging high-resolution assets or multi-gigabyte file handles across monitors.

We rejected Electron entirely. CatchSpace is built as a compiled, native Windows executable. By communicating directly with the Windows Subsystem and Win32 User32 APIs, CatchSpace interacts with the OS window manager at hardware speed, with immediate event delivery and no virtual machine overhead.

2. Event-Driven Hooking vs. CPU Polling

Many naive desktop shelf implementations attempt to track cursor movement and file dragging by setting up a high-frequency polling timer (e.g., executing a loop every 16ms to query cursor position and clipboard states). This is disastrous for laptop battery life, generating continuous CPU wakeups that prevent processor cores from entering deep C-states.

Instead, CatchSpace utilizes an event-driven model based on Windows OLE Drag and Drop interfaces: IDropTarget and Win32 low-level hooks. When a user initiates a drag gesture within Windows Explorer or any desktop application, Windows broadcasts shell notifications.

// Simplified conceptual snippet of OLE IDropTarget implementation HRESULT __stdcall CatchSpaceDropTarget::DragEnter(IDataObject* pDataObj, DWORD grfKeyState, POINTL pt, DWORD* pdwEffect) { if (ContainsValidFileDrop(pDataObj)) { // Trigger hardware-accelerated shelf slide without polling WakeShelfTrigger(pt.x, pt.y); *pdwEffect = DROPEFFECT_COPY | DROPEFFECT_MOVE; return S_OK; } *pdwEffect = DROPEFFECT_NONE; return S_OK; }

When no drag operation is underway, CatchSpace sleeps. It consumes 0.00% CPU cycles and generates zero interrupt wakeups. The shelf only computes geometry when Windows signals that a drag event has crossed the screen boundary threshold.

3. Active Working Set Trimming

RAM consumption on Windows is divided into two key metrics: Committed Memory (virtual memory allocated) and Working Set (physical RAM actively held). Many developers assume that whatever memory is allocated remains locked in physical RAM forever.

CatchSpace employs proactive working set optimization. When the shelf drops back into hidden standby mode after a successful file transfer, the engine signals the Windows memory manager via EmptyWorkingSet() and releases cached UI buffers that are not immediately necessary.

Result: While active during a drag-and-drop transaction, CatchSpace peaks at roughly 35MB–38MB of RAM. Once settled in the background, physical RAM residency drops even lower, allowing your system to allocate memory where it actually belongs: your compilers, 3D engines, and creative tools.

4. Multi-Monitor Geometry & DPI Awareness

One of the hardest challenges in native Windows desktop utility development is Per-Monitor DPI V2 awareness. In a modern setup, developers routinely pair a 4K 150% scaled primary display with a 1080p 100% secondary display.

If an application is not strictly Per-Monitor V2 aware, dragging a file across the monitor seam results in blurry UI scaling, mismatched hitboxes, and dropped OLE payloads. CatchSpace hooks WM_DPICHANGED dynamically:

  1. Calculates physical screen bounds using MonitorFromPoint and GetMonitorInfo.
  2. Dynamically recalibrates dropzone margin offsets to ensure the shelf appears at the exact physiological edge of whichever screen the cursor currently occupies.
  3. Renders vector glyphs and drop surfaces at native resolution without bitmap interpolation.

5. The Philosophy of Invisible Tools

The best tools are those you forget are running. CatchSpace has no onboarding splash screens, no subscription prompts, no background telemetry threads phoning home, and no system tray notification spam. You drop a file on the shelf, maximize your destination window, pull the file out, and the shelf vanishes.

We believe the indie software movement must lead the return to performance-first engineering. When software is crafted with respect for the user's hardware and privacy, everyone wins.

Experience CatchSpace for Yourself

Native, lightweight, and completely free. Available for Windows 10 and Windows 11.