Re: [Fwd: Re: A console: method for accessing standard streams]

[Giovanni Corriga <valkadesh libero it>]

Il lun, 2002-11-18 alle 12:26, Alexander Larsson ha scritto:
> On 18 Nov 2002, Giovanni Corriga wrote:
> 
> > Il lun, 2002-11-18 alle 10:02, Alexander Larsson ha scritto:
> > > 
> > > Also, the uri-scheme is a bit strange. A uri is supposed to be a 
> > > location of an object, not something that differs for each process that  
> > > resolves it.
> > 
> > I don't really follow you here. If two processes get a dynamic HTML page
> > by using the http: method, the content of the page may be different for
> > each process.
> 
> Yeah. Or the file referenced by file: could have changed. But its still in 
> theory a reference to a particular file object or web page, even if the 
> exact bits in the file may change. Referencing the standard input streams 
> is (in my opinion) quite different though. I would much prefer a way to 
> e.g. get a GnomeVFSHandle for an open filedescriptor (unnamed handle, no 
> "fd:2" uri) instead of having magic names like this. 

Yes, but how could you get this handle? You would need some special API
calls just for this. With a console: method, we can use the API we
already have.

>  
> > > And what will nautilus do if someone typed console:///stdin 
> > > in the location bar? Will it even work?
> > > 
> > If you type console:///stdin in Nautilus, you will get an error message
> > telling you that Nautilus doesn't have a suitable viewer.
> 
> So, what is the right behaviour? Should it open the text viewer? When is 
> the URI itself ever useful to a programmer or a user? (As opposed to the 
> gnome vfs handle. I can see that being useful.)
> 
I think the right behaviour is not doing anything, just like when you
try to open a socket or a device file.

	Giovanni