Fix nginx caching

[dfabulich]

Fixes #67

A long summary of `curl` test results

I used a very basic options.json: { "domain": "localhost", "subdomains": true } (with no support_bypass set).

Initially, I tested in curl, like this. (I've trimmed irrelevant lines.) Throughout, I checked the logs to see what the app server returned.

# Home page; has no last-modified header, so nginx returns 200 / Expired
% curl -o /dev/null -v http://localhost
< HTTP/1.1 200 OK
< Cache-Control: max-age=1
< X-Cache: EXPIRED

# Index page; app server returns 304, nginx returns 200 / Revalidated
% curl -o /dev/null -v "http://localhost/?url=https%3A%2F%2Fifarchive.org%2Fif-archive%2Fgames%2Fcompetition2024%2FGames%2FTHE_BAT.zip"
< HTTP/1.1 200 OK
< Cache-Control: max-age=1
< Last-Modified: Thu, 17 Oct 2024 03:22:28 GMT
< X-Cache: REVALIDATED

# Index page conditional get; app server returns 304, nginx returns 304
% curl -o /dev/null -v --header "If-Modified-Since: Thu, 17 Oct 2024 03:22:28 GMT" "http://localhost/?url=https%3A%2F%2Fifarchive.org%2Fif-archive%2Fgames%2Fcompetition2024%2FGames%2FTHE_BAT.zip
< HTTP/1.1 304 Not Modified
< Cache-Control: max-age=1
< Last-Modified: Thu, 17 Oct 2024 03:22:28 GMT
< X-Cache: REVALIDATED

# HTML file on subdomain; app server returns 304, nginx returns 200 / Revalidated
% curl -o /dev/null -v "http://2ckvcbhjia.localhost/2ckvcbhjia/index.html"
< HTTP/1.1 200 OK
< Cache-Control: max-age=1
< Last-Modified: Thu, 17 Oct 2024 03:22:28 GMT
< X-Cache: REVALIDATED

# Conditional get of HTML file on subdomain; app server returns 304, nginx returns 304
% curl -o /dev/null -v --header "If-Modified-Since: Thu, 17 Oct 2024 03:22:28 GMT" "http://2ckvcbhjia.localhost/2ckvcbhjia/index.html"
< HTTP/1.1 304 Not Modified
< Cache-Control: max-age=1
< Last-Modified: Thu, 17 Oct 2024 03:22:28 GMT
< X-Cache: REVALIDATED

# "Safe" CSS file on subdomain, which we want to redirect to main domain
# App server returns 302 with no "Last-Modified" date; nginx returns 302 / Expired
% curl -o /dev/null -v "http://2ckvcbhjia.localhost/2ckvcbhjia/interpreter/style.css"
< HTTP/1.1 302 Found
< Cache-Control: max-age=1
< Location: //localhost/2ckvcbhjia/interpreter/style.css?lastmod=1729135348000
< X-Cache: EXPIRED

# Safe CSS file on main domain with `?lastmod` parameter
# App server initially returns 200 OK with max-age=604800 (one week)
# From then on, nginx skips querying app server, returns 200 / Hit
% curl -o /dev/null -v "http://localhost/2ckvcbhjia/interpreter/style.css?lastmod=1729135348000"
< HTTP/1.1 200 OK
< Cache-Control: max-age=604800
< Last-Modified: Thu, 17 Oct 2024 03:22:28 GMT
< X-Cache: HIT

# Conditional get of CSS file on main domain with `?lastmod` parameter; nginx returns 304 / Hit
% curl -o /dev/null -v --header "If-Modified-Since: Thu, 17 Oct 2024 03:22:28 GMT" "http://localhost/2ckvcbhjia/interpreter/style.css?lastmod=1729135348000"
< HTTP/1.1 304 Not Modified
< Cache-Control: max-age=604800
< Last-Modified: Thu, 17 Oct 2024 03:22:28 GMT
< X-Cache: HIT

# 404 page; nginx always queries app server and returns 404 with max-age=0, no Last-Modified or X-Cache header
% curl -o /dev/null -v "http://localhost/404"
< HTTP/1.1 404 Not Found
< Cache-Control: max-age=0

Finally, I enabled nginx.cache.support_bypass in options.json and repeated this test:

# HTML file on subdomain with client headers to bypass cache
# app server returns 200 (nginx doesn't pass If-Modified-Since), nginx returns 200 / Bypass
% curl -o /dev/null -v --header "Pragma: no-cache" --header "Cache-Control: no-cache" "http://2ckvcbhjia.localhost/2ckvcbhjia/index.html"
< HTTP/1.1 200 OK
< Cache-Control: max-age=1
< Last-Modified: Thu, 17 Oct 2024 03:22:28 GMT
< X-Cache: BYPASS

# Conditional get of HTML file on subdomain with client headers to bypass cache
# app server returns 200 (nginx doesn't pass If-Modified-Since), nginx returns 304 / Bypass
% curl -o /dev/null -v --header "Pragma: no-cache" --header "Cache-Control: no-cache" --header "If-Modified-Since: Thu, 17 Oct 2024 03:22:28 GMT" "http://2ckvcbhjia.localhost/2ckvcbhjia/index.html"
< HTTP/1.1 304 Not Modified
< Cache-Control: max-age=1
< Last-Modified: Thu, 17 Oct 2024 03:22:28 GMT
< X-Cache: BYPASS

Then, I tested "The Bat" in Chrome with http://2ckvcbhjia.localhost/2ckvcbhjia/index.html.

I tested with "Disable cache" checked and support_bypass disabled. I got:

• 200 OK / X-Cache REVALIDATED for the HTML; app server returned 304
• 302 Found / X-Cache EXPIRED for the subdomain subresources
• 200 OK / X-Cache HIT for the main domain subresources with ?lastmod
  • I confirmed that nginx did not contact the app server for these requests
• 200 OK / X-Cache REVALIDATED for Background.jpg, which is referred to by style.css; app server returned 304

This was 2.1 MB transferred for 20 requests.

Then, I refreshed with "Disable cache" unchecked. I got:

• 304 Not Modified / X-Cache REVALIDATED for the HTML; app server returned 304
• 302 Found / X-Cache EXPIRED for the subdomain subresources
• 200 OK (from disk cache) for the main domain subresources with ?lastmod
  • I confirmed that the browser did not contact nginx for these requests
• 304 Not Modified / X-Cache REVALIDATED for Background.jpg, which is referred to by style.css; app server returned 304

This refresh was 3.4kb transferred for 20 requests.

If I enable nginx.cache.support_bypass in options.json and "Disable cache" in Chrome, I find that nginx never passes "If-Modified-Since" headers, and always returns X-Cache: BYPASS.

[dfabulich]

Wanted to ping about this PR. Is it ready to merge?

[curiousdannii]

Sorry, I forgot about it. I'll take a look soon.