Why is the socket interface the way it is? Why isn't it like /dev/tcp in bash?
I tried Plan 9 about 10 years ago. Since then I use VPNs everyday and it can be tricky to setup at times.
The Plan 9 concept of just importing /net from the VPN machine as a union overlay is so elegant.
https://en.wikipedia.org/wiki/Plan_9_from_Bell_Labs#Combinin...
That's wrong. /dev/tcp or /dev/udp should be devices themselves, not a fake one for bash.
Open /dev/tcp/google.com/80 or /dev/tcp/192.168.1.1/22 or whatever.
Instead of having your web server start as root and then do some UID switching to drop priveleges why not have it start as a non priveleged user from the beginning?
The ports under 1024 are reserved for root but how about we get rid of that and have files for every port number? Then use chown and chmod to set permissions and only let the httpd user account read and write to port 80?
Also, rant whatever you want, but importing /net it's far better than all of the IPTABLES NAT crap you have to write in order to have an usable system build on top of stacked crap again and again.
9front is the future, and not that turd based on half baked networked file systems from Sun named NFS, or worse, Java as the 'standard' of enterprise 'computing'.
Don't let me start on X11, Bash having /dev/tcp and so on.
Or having to use VNC instead of X11's forwarding in order to be usable today (Or NX, Xpra...), thus, defying the basic scheme of X11's client/server model.