I'm looking to supply a process that only reads from a local file with content from an http URL. The process is a daemon that is hard-coded to read a specific file in it's current working directory and reads it whenever a particular event happens.

I have no control over the event that triggers the file read and I want to supply near real time data instead of constantly dumping data via a curl script to the file.

What I thought might work would be a fifo that executes a command whenever it is read. This would then be connected to netcat or curl to retrieve the contents of the URL (xml data). Is this possible?

  • If the process is well-behaved, that is, it opens the file, reads it sequentially, and closes it, a named pipe ought to work. Something like while true; do (echo -n ""; curl ... > tmp; cat tmp) > /path/to/fifo; done. Jan 21, 2015 at 23:48
  • See also Create a virtual file that is actually a command. I'm not voting to close as a duplicate because the specific use case of reading from an HTTP URL can have simpler answers than the generic idea of executing commands to implement file reading and writing. Mar 30, 2015 at 12:03

1 Answer 1


It's impossible to make reading from a pipe launch a process on a normal filesystem. But there are other methods that should do what you need.

You can run the content-generating program in a loop. Each time the FIFO is opened for writing, this will block until a reader comes along.

while true; do
  generate-content >fifo

Another approach is to make a FUSE filesystem that generates whatever content you like — the data returned to the reader is generated by the filesystem driver, which is an ordinary program, so you can make it do whatever you like. Building a FUSE filesystem driver is a lot of work though, so unless you find one that already does exactly what you need this is probably overkill. Some possible filesystems that may help you are:

  • ScriptFS — judging by the description, this does exactly what you need: “replicates a local file system, but replaces all files detected as ‘scripts’ by the result of their execution”. I've never used it.
  • HTTPFS — reading from the file triggers a web download; connect it to localhost and run a small webserver such as lighttpd or thttpd that supports CGI or a similar method to serve dynamic content. If the data you need is already downloaded from a webserver, simply point HTTPFS at it.
  • It might not be (easily) possible to react to a fifo read (though flooding a pipe could work, maybe) but it would be possible with a serial line/terminal. screen could be setup to handle this, or SLIP and/or PPP (even SYNC_PPP) could be used - this is kind of what they're for, as I think.
    – mikeserv
    Jan 22, 2015 at 2:32

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .