This proposal has been added to the active column of the proposals project
and will now be reviewed at the weekly proposal review meetings.
— rsc for the proposal review group

Despite the possibility of extended support for customers paying Microsoft, it seems clear that nearly everyone is dropping support for Windows 8, including .NET. We would need to announce in the Go 1.20 release notes and then Go 1.21 would be the first release without support, and then Go 1.22 would mark the end of Go 1.20 being supported, which would be the real end. That Go 1.22 release would be in Feb 2024.

Based on the discussion above, this proposal seems like a likely accept.
— rsc for the proposal review group

proposal: all: end support for Windows 8, Windows Server 2012 in Go 1.21 or 1.22
...
In contrast to Windows 7, 8's corresponding Windows Server releases, Server 2012 and 2012R2, will not fall out of support for some time. Extended support ends later this year, but "extended security updates" don't end until 2026. I don't have a sense of how important that install base is to us.

I'm confused by this wording. Is this saying that Windows 7 (aka Windows Server 2008) will continue to be supported but Windows 8 (aka Windows Server 2012) will not? Only 7 and 10+?

Or is this about unsupporting both 7 and 8?

Or is 7 already unsupported? But https://github.com/golang/go/wiki/MinimumRequirements says:

For Go 1.11 and later: Windows 7 and higher or Windows Server 2008 R2 and higher. We test on Windows Server 2008 R2, 2012 R2, and 2016, which are roughly Windows 7, Windows 8.1, and Windows 10.

As a data point (not even in support of any particular argument), our Go application has Windows user distribution as follows:

Win 10+	 96.8%
Win 7 	 1.7%
Win 8.1	 1.3%
Win 8	 0.1%

Pre-Win10 is low percent, for sure, but in absolute numbers it's still a good number of users. But I agree the Go project can't run builders & support everything forever, especially when upstream doesn't.

In any case, I'll note that https://tip.golang.org/doc/go1.20#ports hasn't been updated yet to say anything about upcoming Windows deprecations. Is that going to happen before rc2/final?

@bradfitz There's a separate proposal for just Windows 7, #57003.

Ah, gotcha. Thanks!

No change in consensus, so accepted. 🎉
This issue now tracks the work of implementing the proposal.
— rsc for the proposal review group

Yeah, I mentioned that in the original comment. Like I said there, the question is how much we're concerned about those users. And as Russ pointed out, it'll continue to get Go security fixes until 1.20 falls out of support in early 2020 2024.

Saw that, after posting and immediately Xd my comment. This is somewhat concerning, but following .NET seems fine to me ultimately so no objection.

Change https://go.dev/cl/457695 mentions this issue: doc/go1.20.html: pre-announce dropping Windows 7, 8, and friends

I'm updating the release notes to announce end of support after 1.20. If there's a good reason to delay to 1.21 we can consider it.

As a data point, I work on the Go datadog-agent and 20% of its Windows install base is still on Windows Server 2012/R2. As a point of comparison, less than 2% is on Windows Server 2008R2.

I can't tell how representative of the Go Windows install base these numbers are, but 20% is a fairly high number.

For what it's worth, the .NET project will continue to support Windows Server 2012/R2 for Windows users who purchase Extended Security Updates. It's likely that a significant number of enterprise Windows users will remain on Windows Server 2012/R2 until closer to the end of the Extended Security Updates in 2026.

I don't know how much of a burden supporting Windows Server 2012/R2 is for the Go project, but is a slightly longer support period still on the table?

Thanks for the information. Do you have a sense of whether those users are likely to pay for the extended security updates? Would 6 months more Go support make a material difference?

In general, we're not going to be able to match Microsoft's paid support timelines, so we have to draw the line somewhere. The end of "extended support" seems like a reasonable place to me. Hopefully that 20% will decline over the next year.

@heschi thanks for the response

Do you have a sense of whether those users are likely to pay for the extended security updates?

We're not seeing a steep downwards trend on the usage of Windows Server 2012 yet, so given the timeline of the end of "extended support" we would expect that quite a few of these users will rely on Microsoft's paid support while they migrate to later Windows versions. But we don't know about the support arrangements they have/plan to have with Microsoft.

Would 6 months more Go support make a material difference?

I think it would to these users, but for sure Go can't match Microsoft's paid support timelines.

Unfortunately I don't have more hard data than this number of 20%. It'd be interesting to see if the announcement in the Go 1.20 release notes of the end of support of Windows 2012 in Go 1.21 will lead to more reactions from the community.

Change https://go.dev/cl/500239 mentions this issue: doc/go1.21: document macOS and Windows requirements

Change https://go.dev/cl/509335 mentions this issue: os: remove 5ms sleep on Windows in (*Process).Wait