Rob Landley wrote in a message to Geert Bosch:

 RL> That's the unix solution, but it creates one heck of a
 RL> nightmarish directory structure...
It's not that bad, I think. We already have \PIPE, \SEM etc directories. So
it seems logical to put devices also in a directory.

 RL> The solution -I- was always fond of was putting a colon
 RL> after the device name.  you can echo ath0 > com1: and
 RL> that'll work fine.
That seems to be a nice solution to. It has the advantage that it doesn't
pollute the name-space for user-files, and they would be easily accessible.

 RL> Also, this way you could consider the various hard drives to
 RL> be "devices".  You could install a ram disk driver and
 RL> access it as ram:\files much more easily.  
Considering OS/2's usage of drive letters, this would be very logical. I've
never comprehended why only a single drive-letter was allowed. 26 devices
may seem a whole lot, but when I've mounted some network devices, I'm
easily running out of them. And I don't have one of those fancy 18-disk
CD-changers.
 RL> Or you could install a print spooler and provide support for 
 RL> something along the line of dir prn: and it could list jobs 
 RL> waiting t be printed.  Sick, eh? 
You mean slick?! The advantage of the system would be the possibility to be
able to see easily what kind of device a certain name refers to. I think
the best solution is to extend the naming model the way you described, and
allow symbolic links to be able to create a stable view of the computers
file-systems, even when files physically change location.

Groetjes, Geert 
--- timEd/2 1.01+