pipe/redirect dumb question

[Erik Price]

On Wednesday, December 11, 2002, at 07:00  PM, jim.mcginness at att.net 
wrote:

> An ordinary redirection opens a file descriptor with regular file 
> semantics.
> Some shells implement "process" redirection which is similar to a pipe.
>
> The main difference between pipe file descriptors and "file" file 
> descriptors
> is that pipes implement a sort of flow control. Draining a pipe causes 
> the
> reader to block rather than receive an EOF. Filling up the pipe will 
> cause
> the writer to block.
>
> Other differences are the way the file descriptors respond to various 
> other
> syscalls. For instance, you can't do a 'seek' on a pipe.

Thanks for this comprehensive answer.  Originally I asked this question 
from a strictly pragmatic approach ("does it make a difference if I 
pipe from a utility or redirect from a file, if it's the same data 
either way"), but I had never considered the way piped data works.  
Specifically that it causes the reader/writer to block depending on the 
circumstances.

Does the code that tells the reader or writer to block have to be 
written into the application, or is that somehow built into the shell?  
If it needs to be built into the application, well then say for 
instance I'm writing a Perl script.  I've never programmed my Perl 
script to block if the writer gets "full".  So perhaps it's handled 
internally by Perl.  But if I'm writing a C program, I shouldn't make 
the assumption that this will be done for me.

Unless of course I'm totally off the mark here, and the rules of piping 
are handled by the shell or OS or something.


Erik





--
Erik Price                                   (zombies roam)

email: erikprice at mac.com
jabber: erikprice at jabber.org