extends article.pug block append head link(rel='canonical' href='https://arav.su/stuff/article/nginx_recipes_and_tips') block article header h2 NGiNX's recipes & tips div.menu a(href='/stuff#articles') Go back to articles list time(datetime='2021-11-15') 15 November 2021 nav h3 Contents ol li #[a(href='#art-1') Introduction] li #[a(href='#art-2') Userdir functionality like Apache's mod_userdir] li #[a(href='#art-3') Note on how NGiNX works with HTTP headers] li #[a(href='#art-4') Tell a Tor visitor of your clearnet site that you have an onion] li #[a(href='#art-5') Variables] h3#art-1 #[a(href='#art-1') 1. Introduction] p Here I'll place recipes for implementing different functionalities and notes on webserver's behaviour. h3#art-2 #[a(href='#art-2') 2. Userdir functionality like Apache's mod_userdir] p Once I wanted to have a “tilde user directories” like #[code /~user/] which is more known as Apache's #[code mod_userdir] feature. I work with NGiNX so regular expressions is the way to do that. p Nothing special in my case. I only need to keep files there, so others may get them. Thing is simple, but I took some time to realise what regexp to use here back then. p The implementation is quite simple: pre | location ~ ^/~(.+?)(/.*)?$ { | alias /home/$1/pub$2; | autoindex on; | } p Of course, you can choose whatever place for user's public directories. In my case you need to give a read and execute permissions to user's home directory and to public directories inside them to others (#[code chmod o=rX /home/user]), which may be a security concern. p #[code autoindex on] will make an index of files that lies by URL. And, of course, you can put there an #[code index.html] file. h3#art-3 #[a(href='#art-3') 3. Note on how NGiNX works with HTTP headers] p Let's say in #[code http] block you specified common headers like #[code X-Frame-Options], #[code X-XSS-Protection], and so on for all #[code server] directives to use. But, if you add some other header for a specific #[code server] or #[code location] block then all those headers would be dropped. p For now the only cure for it is to place all that headers in a separate file like #[code common-headers.inc] and using #[code include] directive to include them in all the #[code server] and #[code location] blocks where additional headers are added. h3#art-4 #[a(href='#art-4') 4. Tell a Tor visitor of your clearnet site that you have an onion] p There is a custom HTTP header #[code Onion-Location] that is being recognised by Tor Browser and it will show a ".onion available" button in an address bar. p The code for NGiNX is this simple: pre | add_header Onion-Location http://yoursite.onion$request_uri; h3#art-5 #[a(href='#art-5') 5. Variables] p Just wanna tell about this cool feature you can use to reduce code repetition. Alas, you can use them only in #[code server], #[code location] and #[code if] blocks. If only they could be declared in a #[code http] block, they'd be perfect. But still, they help a lot. p I heavily use aliases pointing to directories within the same base directory. p Judging from a little testing, variables doesn't decrease performance. Or I just don't notice it. p Variables could only store strings. To demonstrate the usage I'll just post here a part of my config: pre | set $root /srv/http/dwelling; | ... | location =/favicon.ico { | alias $root/shared/static/img/favicon.ico; | } | ... | location /assets/ { | alias $root/homepage/static/assets/; | } p The very reason why I had to use variables is that #[code alias] doesn't work with relative paths. I wish #[code alias] could be relative to #[code root], but we are forced to look for workarounds. p Update: Oh, I realised that you can declare variables in #[code http] block using #[code map] directive with #[code default] set to value you need. Funny, I've been using #[code map] to prevent LAN IPs from being logged all this time, heh. So, here's example: pre | map 1 $root { | default /srv/http/dwelling; | }