[ale] Object Model on Linux...and GUI

[R I Feigenblatt]

> I would really like to also start a discussion on the linux user
> interface, or lack thereof. Do you like it how it is, or what would you
> do different? Let's dream a little.

Speaking of interfaces, I'd like to raise the issue of
developments in the X Window System and interoperability
with the world beyond X, especially as concerns Linux.

X was created in the mid-80's to enable endusers to operate
applications executing on other machines (anywhere in the world)
from the I/O devices on their local machine. For the most part,
however, it has been limited to Unix machines (I know about WinDD)
and the TCP/IP protocol (yes, DECNet also).

In the meantime, the early 90's saw the growth of the
World Wide Web, exploding into a mass phenomena by the mid-90's.
Double-digit perecentages of US households now use the Web on at
least a monthly basis; I doubt even 1% of households have ever
even HEARD of X.

X chauvanists laughed at the simple distributed hyperlinking
model of the original Web. With the addition of graphics and CGI
programs, simple interactivity was enabled. In the last year
Java and other technologies for making the Web truly interactive
have emerged. By being a thin entering wedge that delivered the
most function for the least effort first, the Web has changed
the world. The Internet would not be a mass phenomena now were
it not for the "GUI" that the Web put on it, much in the same
way the Mac and Windows made the PC accessible to non-nerds.

In recognition of this, the keepers of the X flame in Cambridge,
the X Consortium, have put forth "Broadway" (which will be the next
release of X in place of X11R7). You can read a bit about it at:
http://www.x.org/consortium/broadway.html

Recognizing that Web browsers are places where X has never made
or expanded a foothold, Broadway provides the ability to run
APP-server-executed applications directly from a Web browser.
Yes, mighty X as just another browser plug-in.

But even this might not "save" X. If Xophiles had at first
scorned the stunted interactivity model of the original Web,
X itself did not lack critics. In "The UNIX-HATERS Handbook",
(1994), written by a trio including two MIT grads who then
worked for Microsoft & Apple, respectively, an entire chapter
is reserved for roasting X. One key complaint is that it is too
simple, in that it assumes no processing other than I/O takes
place on the enduser's machine. Of course, processes running
on different machines (e.g. the enduser's, and some remote
machine) can share access to the local X server and its windows
(a potentially nasty security issue, by the way), but this
coordination might best go beyond X per se.

While Apple is now on its back, seeming to have caught engineering
palookavillitis from its intimate congresses with IBM, (Taligent
and Kaleida, RIP; PowerPC next?) Microsoft is addressing the question
of client-server computing with its Distributed Common Object Model.
Within the Web world per se, Microsoft is reaching BEYOND the
Windows platform by delivering an SDK for creating ActiveX widgets
as native Mac binaries, with Unix to follow, one wishes soon.
With the merging of Internet Explorer (Web browser) and Explorer
(the Windows file manager) in the Nashville release, it is clear
Web standards will be a driving force of the Microsoft Windows GUI.
And with Netscape embracing ActiveX in a stunning turn-around, it is
clear that Windows components will play a big role in the Web's future.

While a breakthrough in its time, X has been a technical laggard in
recent years. One thinks of things like audio, image (vs. graphics)
support and compression. Designed for the wideband ethernet environment,
X did not adapt as well to the thin dial-up phone lines in wide use as
did the Web protocol. Broadway seeks to address this with X.Fast.
 
One disturbing issue to those of us interested in free Unix flavors
like Linux and BSD, is that the new plans for Broadway do NOT extend
to the XFree86 server on which these users depend, according to an
article in the January 1997 edition of "Unix Review". This is especially
troubling as the non-profit X Consortium will fold up shop and give
responsibility for X to The Open Group (successors to X/Open and OSF,
the people who brought us Motif, CDE, DCE, etc.) It might emerge that
people on Wintel and Mac machines finally(*) will be able to use X apps
"for free" if the apps execute under Solaris, SCO, AIX, etc., but not
Linux, should the "X client" libraries available for Linux not support
Broadway features like X.Fast!        (* One has been able to buy
(often flakey) X servers for such machines, but they have seen little
success; a free Web browser plug-in could be a different matter.)

Make no mistake about the importance of Microsoft Windows, despite
the slow start of stuff like NT. Dataquest says 1995 NT shipments
were 2.2 million, and suggests 8 million sales in 1996. IDC says
NT Server outsells Netware 3 or Netware 4 or ALL versions of Unix
combined. Unix has a strong foothold at the very high end, but many
buyers find Wintel hardware using NT much more cost-effective than
comparable traditional UNIX solutions. And the onerous cost of RAM
that once allowed RAM-hungry Windows NT to be laughed at by friends
of RAM-parsimonious Linux is no more! How long until Cracker-Jack
boxes contain SIMMs? ;-)

On this mailing list I have often played the wet blanket, discouraging
unruly noise makers. But now I think it is important that Linux lovers
make sure that it can do well in a changing world. A good first step
might be to assure the Broadway-compliance of Linux X apps, so they
can be accessed from a plethora of non-Linux computers in the future.
(One ASSUMES availability of a Broadway plug-in for Linux web browsers, 
 so if we sit at a machine running Linux we'll be able to have dial-up
 access to apps from mainline Unix boxes. But too bad one might not be
 able to use X.Fast directly vs. X.Fast on Web browser on XFree86!)
Of course, as far as Linux as an app-user (vs. app-server) machine is
concerned, Microsoft would no doubt like someone to write an ActiveX
SDK for it, if something like that is not already subcontracted out.
 
Ron Feigenblatt

P.S. One might imagine that a much-improved Java AWT might one day
surplant, rather than just sit on, X and MS Windows, but not soon.