[Sorry for the split reply, I pressed Send too soon...] Nadav Har'El wrote on 2003-06-30:
> On Mon, Jun 30, 2003, Beni Cherniavsky wrote about "[OT?] Printing on X (was: Re: > [Jog Offer] SGI is looking for a Xwindows hacker)": > > Since I never wrote a program that prints something, I take no stance > > on whether directly emitting PostScript or using the windowing > > system's calls is an easier way to program printing. NeXTStep comes > > The reason I wanted printing support in the ICCCM was not because I > wanted to avoid Postscript (in fact, I like Postscript). The issue that > bothered me was that the basic idea behind X was that the applications > I am running can run on different machines and communicate through the > X server. When one application wants to print (and similary, play sounds) > it should not use that remote's machine printer, or the "PRINTER" environment > variable on the remote machine, but rather it should have sent the file to > print to some "Printer Manager" through the standard X-Windows communication > mechanisms. > > People that say "X communication is slow", "we better have direct TCP > communication", "there's the LPR protocol" forget that not only is setting > up direct communication a mess (you need to seperately configure every > account you have), in many cases it is completely impossible because the > seperate machines have no TCP/IP connection. The "modern" example to think > of is SSH's X forwarding - when you have an X connection but cannot make > direct TCP connections because of firewalls. > Now I see what you meant, and I agree. Moreover, I think this is a fundamental mistake in the X architecture - they didn't think that people will want to have many related services network-transparent and tied to a "desktop". Similar situation happens with esd/artsd. I think they should have split X into a separate "Desktop Manager" that handles authentication and an extensible set of forwarded ports and the "Graphics Server" that handles graphics and probable keyboard/mouse and is just one of the forwarded ports. That way, many X extensions that are/were held back by the need to integrate with the X wire protocol and by the fact that conceptually they shouldn't be part of X at all, would be much easier. In fact, now with SSH we do have most prerequisites for such an architecture. The missing part is an easy way to register interesting services with SSH to be forwarded. -- Beni Cherniavsky <[EMAIL PROTECTED]> ================================================================= To unsubscribe, send mail to [EMAIL PROTECTED] with the word "unsubscribe" in the message body, e.g., run the command echo unsubscribe | mail [EMAIL PROTECTED]
