Why do we need a constant that has the same value as the zero value for that type? Why does explicitly saying "the thing isn't set" improve over just not setting it?

Change https://golang.org/cl/237998 mentions this issue: net/http: add SameSiteUnsetMode const

@davecheney In my opinion it mainly improves code readability in some situations. The behavior is already there, but the name is not, so to invoke such behavior I need to pass 0.

For example you can have a configuration with SameSite mode and you don't want to have multiple ways to construct a cookie (with or without the samesite attribute). For example see the ory/hydra#1908 commit.

Thanks for the example. That API is pretty hard to use correctly, especially as there is no safe default for that parameter. Changing Go might be a solution to that, but there might be others which would lead to a more usable API.

@davecheney Agreed, the cookie samesite API is hard to use and one of the small problems is that there is behavior that does not have a name, yet it is used. The other is that SameSiteDefaultMode is a name confusing people, as the mode is not a safe default, nor default, nor safe. The first one is easy to fix, but I'm not sure how to fix the other one without breaking the compatibility. You suggest there might be other ways to enhance the API, can you be more specific?

CC @bradfitz @neild

I'm a bit confused about SameSiteUnsetMode vs SameSiteDefaultMode vs SameSiteNoneMode.
(I know what they mean, I just find the names confusing.)
It seems like leaving 0 for "there's just no SameSite at all" might be clearer.

Also, based on the change made for #36990, the proposed SameSiteUnsetMode would be exactly the same as SameSiteDefaultMode, which is even more confusing.

It's unfortunate that we now have 0 and 1 meaning the same thing, but since 1 has a name, it seems like we should leave it there and not name 0 as well. If you want to be explicit about nothing happening, you can use the named 1 instead of a named 0.

@rsc Yes, the names are confusing, I'm currently discussing the changes for #36990. @FiloSottile proposed instead of SameSiteUnset instead of SameSiteUnsetMode and I think it's a step in the right direction, helping to distinguish the zero value from samesite modes.

The idea is to revert the #36990 change and deprecate the SameSiteDefaultMode instead so the user who rely on the old spec behavior are not surprised by the change. Default, None and Unset would be different thing.

I'm still confused about the state of #36990. Until that gets reverted, it doesn't make sense to have two different named constants with the same behavior. Let's wait on this until #36990 settles.

Closing, as the #36990 was settled by not reverting.