Python and other tongues

nginx

multilevel using Node's core net module with rcp streams.

ØMQ's node binding, one among a universe.

as a backend to almost any HTTP server

If we observe controlled TCP/IP loopback locally or published redis project benchmarks, the fact is a unix domain socket does 50% more throughput than a TCP port, see Factors impacting Redis performance, 3rd bullet point from bottom of that section.

Sockets lack overhead, like complex data checking algorithms TCP implements to ensure reliable transport.

Also instead of universally identifiable numberings, Unix domain sockets use the file system as their address name space. They are referenced by processes as inodes in the file system. This allows two processes to open the same socket in order to communicate. However, communication occurs entirely within the operating system kernel.

In addition to sending data, processes may send file descriptors across a Unix domain socket connection.

Of course, us userlanders generally pass a port number among other initializing arguments to Node's HTTP listener.

We know how this runs very fast/reliable TCP port numbered servers. However, Ryan Dahl, Isaacs and core provided an equally trivial mechanism for initializing the runtime via unix socket. Just pass a string directory path in leu of a port number to Node's HTTP function.

A domain communication socket opens and remains at the end of our path.

##any limitations? not really.

The benefit of expanded throughput is measurable latency reduction across the stack surrounding central application transport logic.

####Developers tend to avoid sockets; they are not mainstream.

Although nothing really, here a devil's advocate: unix domain sockets incur a potential cost of additional configuration in meeting the requirements of meaningful process initialization for I/O control. For example, we know a TCP port is available to the browser or request/response interface immediately on initializing.

Unix Domain Sockets do not lend themselves to our common TCP port number req/res object exchange without an upstream mechansim available to pass IO routines from kernel to transport layer, such as Nginx's upstream proxy. Or just another local process.

To open unix style sockets, user permission at the file context is sufficient, unless the upstream proxy passes to port 80.

Port 80 represents an aditional step to account for in meeting its root permission burden during configuration. While this is not unique to TCP port number configuration at the proxy level, once the socket is closed after having established connection to port 80, root permission becomes necessary to remove the closed socket remaining on the file system, though BSD systems may ignore permissions for UNIX domain sockets.

Every time a user's server restarts, it must unlink or delete a potentially root permission locked socket descriptor left behind after socket communcation.

Also the node process must be intialized with root permission in order to write during communication with any proxy passing to port 80, or else it will fail to write a response once an incoming request over 80 has been written to the socket by the root user.