fix: add Vary header to cache interceptor

[sommeeeer]

Closes opennextjs/opennextjs-cloudflare#750

There is some issue with Firefox and possibly some CDNs that they will incorrectly be reusing RSC responses during bfcache restores. Adding a Vary header seems to solve the problem.

[changeset-bot]

🦋 Changeset detected

Latest commit: 64ee772

The changes in this PR will be included in the next version bump.

This PR includes changesets to release 3 packages

Name	Type
@opennextjs/aws	Patch
app-pages-router	Patch
app-router	Patch

Not sure what this means? Click here to learn what changesets are.

Click here if you're a maintainer who wants to add another changeset to this PR

[pkg-pr-new]

Open in StackBlitz

pnpm add https://pkg.pr.new/@opennextjs/aws@916

commit: 64ee772

[sommeeeer]

Thanks for the reviews @vicb and @khuezy

[khuezy]

LGTM. I thought I saw something about nextjs putting the vary header back in the newer release? Do we need to patch it on a specific version?

[sommeeeer]

I thought I saw something about nextjs putting the vary header back in the newer release? Do we need to patch it on a specific version?

Yeah they did, but they added it back in version 15.3.3. I guess theres no point making a patch for NextServer, as we can probably just tell the users to upgrade to that version or higher.

Edit: This is also a really hard edge case to hit only in Firefox, and it does require you to have a redirect in the middleware to a route with RSC. Firefox caches the _rsc response instead of the actual page then.

[vicb]

I thought I saw something about nextjs putting the vary header back in the newer release? Do we need to patch it on a specific version?

Yeah they did, but they added it back in version 15.3.3. I guess theres no point making a patch for NextServer, as we can probably just tell the users to upgrade to that version or higher.

Edit: This is also a really hard edge case to hit only in Firefox, and it does require you to have a redirect in the middleware to a route with RSC. Firefox caches the _rsc response instead of the actual page then.

The reason why I did not formally approve is that I have no clue if the value for the vary header will fix all cases.
If Next uses the same value, maybe you could add a link to the PR/code in the comment?

[khuezy]

We could expand the playwright tests to include the other browsers too.