Why 152.0.5 win64 installer SIGNIFICANTLY larger than win32 installer? >115.16esr on Windows 10 warning?
Up until version 141, the difference between 32 bit and 64 bit Firefox installers wasn't noteworthy, though it has clearly been slowly increasing overtime. However, starting from version 141, the difference between them doubled from 5MB to 10MB, meaning a nearly 13.5% larger installer for the 64bit version than the 32bit one.
Why is that? What happened? Are there are feature differences that would warranty such a large gulf in their installer sizes? It's the AI isn't it? 32bit users were spared the bloat of the AI.
I also see that starting from 115.16 Firefox suggests users that aren't on Windows 7-8.1 or MacOS 10.12-10.14 should use other versions. Given both x86 and Windows' immense reputation surrounding backwards compatibility, is this really a concern? Is there REALLY something supported by Windows 7-8.1 that wouldn't function on 10?
All Replies (5)
Hi
In the last three years, Firefox has gained quite a few new features and improvements (not all of them AI related) that may have increased the size of the installer.
The older versions of Windows and macOS that you mention are noonger supported by Microsoft and Apple. Firefox ESR 115 is provided purely to help people to keep a basic level of web access and updating to a more recent supported version of those operating systems is strongly recommended.
I also see that starting from 115.16 Firefox suggests users that aren't on Windows 7-8.1 or MacOS 10.12-10.14 should use other versions. Given both x86 and Windows' immense reputation surrounding backwards compatibility, is this really a concern? Is there REALLY something supported by Windows 7-8.1 that wouldn't function on 10?
Firefox 116.0 and later requires Windows 10, 11 and macOS 10.15 or later to run.
The older Firefox 115 ESR based on the old Firefox 115 Release was originally going to end as of Fx 115.15.0esr but then Mozilla decoded to give the EOL Windows 7, 8, 8.1 and macOS 10.12, 10.13, 10.14 OS users a web browser that was still secure to use as Chrome/Chromium browsers ended support in early 2023.
Neither of these answer either of the questions I had.
We have tried to address the size issue and the question around support for unsupported operating systems.
How can we help you with Firefox?
Well that's the beauty of open–source, you can use the public tools to inspect the various factors and timelines and charts, and make an opinion of your own. I'll interpret some of that:
Up until version 141, the difference between 32 bit and 64 bit Firefox installers wasn't noteworthy […] starting from version 141, the difference between them doubled from 5MB to 10MB, Why is that? What happened?
WebGPU was added. Both installer sizes jumped by ~5MB for DirectX DLLs bugzilla.mozilla.org/1926651 (security reasons shipping own customized version of these). Later in the cycle it seems the 32bit builds got smaller again, possibly by stripping out this functionality and not having to ship that massive DXC shader compiler thing (I'm not an expert on Win32 dll or graphics so I don't even know if WebGPU works on 32bit FX installs or not — I can only see both installers inflating based on that ticket, and then couple weeks later the 32bit ones getting back smaller; did not trace an exact decision timeline for that, might have been exclusion rules further towards release.)
So that's basically trading off performance and optimization for storage and transport.
I also see that starting from 115.16 Firefox suggests users that aren't on Windows 7-8.1 or MacOS 10.12-10.14 should use other versions.
Because the 115esr is ONLY officially supported for the users on systems that can't upgrade further. If you're not on one of those systems, hence can upgrade to currently supported versions, running this unsupported ESR means you're on your own; and if you want any expectation of support (websites, compatibility, system integration, security), that messaging is to remind you of that.
Given both x86 and Windows' immense reputation surrounding backwards compatibility, is this really a concern?
This has nothing to do with architecture. You can run 140esr or 153 in 32bit if you like. The message correctly conveys just end–of–life specifics between OSes, not technical limitation or changes of the underlying OS.
Is there REALLY something supported by Windows 7-8.1 that wouldn't function on 10?
You might have confused the reasons based on the arch/bitness snippet above. There's a difference between "wouldn't function" and "has been fully QAed by the company as any other supported release" — it's just none of the 115 builds are getting any current automation outside of W7–8.1/OSX10.12–10.14 so they're not tested, measured and quality checked by the tooling and QE staff on OSes they're no longer supported on — as the focus needs to be on the currently supported trains.
Yes, some 115 users can still get 115 updates even though they really shouldn't, but as long as these are being built and offered, even on (a subset of) systems outside of the support profile (list of which is getting shorter over time, as build pipelines are shutting down; the most recent is WinARM64 for 115), anyone can use them as they wish — they're just on their own (read: any perf, compat, platform, stability, security, feature or crash reports would just end up inactionable).