Path: utzoo!utgpu!water!watmath!uunet!lll-winken!lll-tis!pacbell!att-ih!ihnp4!ihopa!ihop3!ihtlt!kosman!kevin
From: kevin@kosman.UUCP (Kevin O'Gorman)
Newsgroups: unix-pc.general
Subject: Re: named pipes
Message-ID: <364@kosman.UUCP>
Date: 14 Apr 88 20:39:19 GMT
References: <339@bacchus>
Reply-To: kevin@kosman.UUCP (Root)
Organization: K.O.'s Manor - Vital Computer Systems, Oxnard, CA 93035
Lines: 22

In article <339@bacchus> darren@bacchus (Darren Friedlein) writes:
>
>OK - Is there anyone out there that can explain to me how named pipes work
>(specifically on the UNIXpc)?  I understand the concept, but I don't know
>how to implement them.  I believe /dev/lp is one, but I'm not sure.  Does
>each one require a device driver?  Does the fact that I use the pty package
>make things any easier?

The only named pipe in the standard UNIX PC stuff is /usr/spool/lp/FIFO, which
I have guessed (but don't know) is how 'lp' tells the scheduler there's stuff
to print.  This exists only when the lp scheduler is running, or has crashed.

I found this out doing backups, because cpio would choke on a named pipe in
the file system.  There's not much documentation on these beasts, but there
was a posting a long time ago called '2-d pipes' which used these things to
create a more complicated pipeline structure than the straight line usually
used.

The idea is to make a directory entry with mknod(2), and then have two
processes open it, one for reading and one for writing.  They are now connected
by a pipe, just like the one you get from the pipe(2) call.  They are useful
primarily when the name of the pipe is agreed by convention (as is the case
with /usr/spool/lp/FIFO).