aboutsummaryrefslogtreecommitdiffstatshomepage
diff options
context:
space:
mode:
authorEgor Tensin <Egor.Tensin@gmail.com>2021-03-04 02:50:38 +0300
committerEgor Tensin <Egor.Tensin@gmail.com>2021-03-04 02:56:06 +0300
commit07beafa9de8f7740bb6db9244c251a2e1d574e84 (patch)
treecc306bf76015fe8b4db02a0984abaaf80f1a6368
parentpacman: add dependency on procps-ng (diff)
downloadlinux-status-07beafa9de8f7740bb6db9244c251a2e1d574e84.tar.gz
linux-status-07beafa9de8f7740bb6db9244c251a2e1d574e84.zip
update README, add some comments, etc.
Diffstat (limited to '')
-rw-r--r--README.md2
-rwxr-xr-xapp.py34
-rw-r--r--img/example.pngbin154692 -> 171835 bytes
-rwxr-xr-xserver.py3
4 files changed, 38 insertions, 1 deletions
diff --git a/README.md b/README.md
index 745cf83..23d3161 100644
--- a/README.md
+++ b/README.md
@@ -17,7 +17,7 @@ Usage
To start a web server on port 18101:
- > python3 -m http.server --cgi 18101
+ > python3 server.py
There's also a [systemd unit] that you can adjust to run the server.
diff --git a/app.py b/app.py
index 020d55a..96ffc1d 100755
--- a/app.py
+++ b/app.py
@@ -5,6 +5,40 @@
# For details, see https://github.com/egor-tensin/linux-status.
# Distributed under the MIT License.
+# Initially it was just index.html and a bunch of shell scripts in cgi-bin.
+# The web server could be launched using just `python -m http.server --cgi`,
+# and all was well. Until I decided that I want to package this little app
+# of mine so that it can be easily installed and run on a variety of systems.
+# There was a problem with that, however: the app was designed to be run by
+# a regular user, not root (mostly due to all the `systemctl --user` calls),
+# and it didn't play nicely with the whole package idea. Plus, I've long
+# wanted to make it show all the systemd user instances, not just the running
+# user's. In addition, I wanted to dedupe the small chunks of code in cgi-bin.
+#
+# So this script was born. It could handle a lot more complexity than the
+# shell scripts; it could be run as root and as a regular user, etc. It was
+# still a CGI script (and it still is BTW), and a separate Python instance was
+# used by the web server to execute it.
+#
+# There was a problem, of course. The thing is, I for once had a very strict
+# performance requirement for this script: it had to work nicely on my rather
+# old Raspberry Pi 1 Model B+. It, being a Python script, didn't. After
+# some investigation, it was obvious that the main problem were the imports
+# (especially those of the cgi & logging modules); they would take more time
+# than the interval between requests. So I had to find a way to make it more
+# performant.
+#
+# I played with Lua for a bit, but couldn't quite make it work in a
+# satisfactory way. Then I had an idea: I could use http.server's HTTP server
+# classes, overriding them to handle the "CGI" requests using the code in this
+# script. That way, the imports would only be done once for the entire life
+# of the server. In addition, no more tricks with sudo were required
+# (http.server runs CGI scripts as user nobody, which is a no-no for this app;
+# some trickery like whitelisting nobody in /etc/sudoers.d were employed).
+#
+# So that's where we are now, and it works! Doesn't load my good ol' Raspberry
+# no more.
+
import abc
import cgi
from collections import namedtuple
diff --git a/img/example.png b/img/example.png
index 0e5a54c..a2cb95c 100644
--- a/img/example.png
+++ b/img/example.png
Binary files differ
diff --git a/server.py b/server.py
index a76d2ca..35099cf 100755
--- a/server.py
+++ b/server.py
@@ -5,6 +5,9 @@
# For details, see https://github.com/egor-tensin/linux-status.
# Distributed under the MIT License.
+# This script launches a HTTP server and uses app.py for processing a set of
+# custom URLs. See that file for the reasons behind this.
+
import http.server
from app import Request