Paper conversion frontend

Server configuration

Create a relatively unprivileged nodewww user to run the server process:

sudo useradd -m nodewww
sudo passwd nodewww
sudo usermod -a -G sudo nodewww
sudo usermod -s /bin/bash nodewww

Enable key-based SSH login for that user:

sudo su nodewww
mkdir ~/.ssh
vim ~/.ssh/authorized_keys  # And paste in your public key here
chmod 700 ~/.ssh
chmod 600 ~/.ssh/authorized_keys

Disable password-based SSH login for security (make sure a key is working first!):

sudo vim /etc/ssh/sshd_config  # And set: PasswordAuthentication no
systemctl restart ssh

Assuming you are running an Ubuntu/Debian-like server, install the required packages like this:

sudo apt install libreoffice-nogui inkscape npm nginx gcc make ruby-dev
sudo gem install anystyle-cli

Now install a new, full version of TexLive, at least 2023 or newer, which may not be available via the package manager. It is easier to take a more manual approach. Get the install-tl-unx.tar.gz file from a recent version. Then run (with adjustments for "2023" as needed):

tar xf install-tl-unx.tar.gz
cd install-tl-20*
sudo ./install-tl  # Press "I" when prompted
echo 'export PATH="/usr/local/texlive/2023/bin/x86_64-linux:$PATH"' >> ~/.bashrc
export PATH="/usr/local/texlive/2023/bin/x86_64-linux:$PATH"
make4ht --version  # Verify new version comes up

Then install conda (as the user that will run NodeJS) if you don't already have it:

su nodewww
cd
wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh
sh Miniconda3-latest-Linux-x86_64.sh

Likewise install n, a NodeJS version manager, as that same user. n allows you to run a specific version of NodeJS:

sudo npm install -g n
sudo n 19  # Install NodeJS version 19.x

Fix problem with ImageMagick refusing to convert PDFs because of an old security problem:

sudo vim /etc/ImageMagick-6/policy.xml
# Look for <!-- disable ghostscript format types --> and delete all of those restrictions

Close and reopen the terminal or run bash to start a new one and get everything in the path.

Installation

Clone this repository onto the server, e.g., into the home directory of the user that will run the server.

Also clone the paper-convert-scripts repository and complete the setup for that. If you installed it somewhere other than alongside this repository, you will need to edit package.json (for this repository, not that one) to point conversion_scripts_dir to the correct place. Also set admin_page_password to something memorable (security provisions are super basic, and not intended to defend against malicious parties so much as inquisitive ones).

Then install the exact version of everything from package-lock.json by running:

npm ci

Set up SSL certificates

sudo cp paper_nginx /etc/nginx/sites-available/
sudo ln -s /etc/nginx/sites-available/paper_nginx /etc/nginx/sites-enabled/
# Now replace the REPLACE_WITH_DOMAIN part with your domain, e.g., papers.example.com
sudo vim /etc/nginx/sites-available/paper_nginx
sudo rm /etc/nginx/sites-enabled/default  # Remove default config
sudo systemctl restart nginx
sudo snap install core
sudo snap refresh core
sudo snap install --classic certbot
sudo certbot --nginx

At this point you should probably install security updates and restart your server:

sudo apt upgrade
sudo shutdown -r now

Running the server

From the paper-convert-www directory, as your unprivileged user, run:

conda activate paper_convert
./restart.sh

If it does not say "Listening on port 3000", something probably did not work. Check ps aux | grep 'node index' to see if it is running, and, if not, npm start to get all possible error output.

Backing up/downloading papers

Note the trailing slashes are important!

rsync -rtuvm --include="*/" --include="*.zip" --exclude="*tmp.docx" --include="*.docx" --exclude="*" nodewww@SERVER:paper-convert-www/papers/ ~/Downloads/pcpapers

# --include/--exclude stuff can be dropped to get everything (not only zip files)
# r = recursive
# t = send timestamps?
# u = only update files with newer timestamps
# v = verbose (show files sent)
# m = avoid creating empty directories in the destination
# Add --dry-run if needed to see what would be transferred

Deleting old papers

After backing up papers, it might be wise to free up disk space by deleting old papers. You can do do this from the paper-convert-www/papers directory:

# See what will be deleted (older than 60 days, for example)
find . -maxdepth 1 -type d -mtime +60 -print
# Delete, only after being sure it is a good idea
find . -maxdepth 1 -type d -mtime +60 -exec rm -rf {} \;