Time to drop 32-bit Windows?

Dear Windows users,

I wonder whether the time has come to drop the distribution of binaries for 32-bit Windows. It gets slowly harder to support. The Java binding is compiled against the pretty old OpenJDK-14, there is no 32-bit Python binary available (at least, not from the official download page).

Does anyone has insight whether there is still demand for 32-bit Windows versions?

Note that the somewhat longer term plan is to drop 32-bit support completely. At the moment it is used for Windows-32, WASM and with Raspbian. Raspbian seems to be slowly replaced by 64-bit Raspberry Pi OS. 32-bit Windows is dying and WASM can hopefully be switched to 64-bit in not too long. Data representation can be seriously simplified by moving to 64-bit-only, removing some restrictions and gaining some performance. Anyone who seems problems with that?

1 Like

My view would be to create one more version with the Windows binaries but not release them. Then if users need them they will notify you, you will know and get an understanding, and then make them public if a user really needs them.

AFAIK Windows 11 does not even come in a 32-bit version and Windows 10 which does have a 32-bit version is nearing the end of life on October 14, 2025.

1 Like

Hi Jan!

macOS dropped 32-bit support some time ago. I agree with you that it is convenient to focus on 64-bits (or at least, keep 32-bits for toy or legacy applications).

Unfortunately, I do not see that wasm64 (WebAssembly in 64-bits) will happen soon in browsers (probably it works in some server-side wasm, which some people see as an alternative to containers).

I’ve found that this issue that has been created recently:
Tracking issue for phase 4 advancement · Issue #43 · WebAssembly/memory64 · GitHub

I have no idea how we can help to make it happen, but at least it’d be convenient to subscribe and react to show that this is important for some developers (who implement dynamic languages with tagged words).

I also found Feature Extensions - WebAssembly, which sort of suggest there is progress.

A post was split to a new topic: Should SWI-Prolog implement an HTTP server?