New comment by atweiden on void-packages repository

https://github.com/void-linux/void-packages/pull/39225#issuecomment-1244705605

Comment:
> You're trying to enable this experimental feature on for everyone, not an opt-in for someone.

Perhaps I’ve miscommunicated.

`--enable-libsodium` at compile time doesn’t activate `cryptmethod=xchacha20` at runtime.

Vim’s `cryptmethod` is a user setting. This setting is disabled by default.

All `--enable-libsodium` does at compile time is allow users to, at runtime, *opt in* to using xchacha20+poly1305 to encrypt a file via running `:set cryptmethod=xchacha20`. This is as opposed to users being limited to running e.g. `:set cryptmethod=blowfish2`. Neither of these commands is ever run in the course of normal Vim operation.

> I (personally) don't want to see bug report for vim `x.y.z` can't read some binaries produced by vim `a.b.c`.

In the worst case scenario, users can 1) compile a specific version of Vim before compatibility was broken either through `xbps-src` or manually, 2) decrypt the affected files, 3) re-install Vim via `xbps-install` if necessary, and 4) re-encrypt the affected files with the new version of Vim. This is relatively easy to accomplish, and would only affect users who elected to run `:set cryptmethod=xchacha20`.

That may sound like a lot, but Vim compiles relatively quickly and easily, and the feature comes with a clear warning in its documentation. If Vim’s xchacha20 implementation breaks between versions upstream, it wouldn’t be a valid reason to veer from the existing void-packages strategy of steadily updating the Vim template.