Alan Kay <Alan.Kay at squeakland.org> wrote:
> Folks --
> 
> This has always flabbergasted me. Why isn't the standard print port 
> on current day OS's represented at least by a pseudofile called 
> "Print". Even if they "don't know nothin' about objects", this was a 
> technique that goes all the way back to the idea of streams as 
> articulated by Chris Strachey in the 60s for !@#$%^ sake! In several 
> of his OS designs (one of which got implemented at PARC), many things 
> were mapped to stream protocols, including regular files, and much 
> other IO.
> 
> What is so !@#$%^& difficult about this idea in this day and age? (or 
> any day and age?)


On Unix, the difficulty is with user code versus kernel.  Unix has this
wonderful idea that most things can be placed in the filesystem.  But it
has this horrible idea that only special kernel code can achieve it. 
Who wants to get ghostscript working in kernel mode??

Incidentally, the GNU Hurd doesn't have that limitation, as I understand
it; you can write kernel extensions in user space, and get all the
privilages that user code normally gets (files, virtual memory,
networking, etc.)


-Lex