% fortune -ae paul murphy

Why haven't GUIs progressed?

My wife asked a variant on this question the other day, particularly with respect to shortcuts that let you bypass the user friendly stuff once you outgrow it.

Sadly I haven't much of a clue here. It seems clear that the Apple GUI has consistently been the most effective among current products in part because Apple split the GUI (including its shortcuts) between applications and operating system to enforce consistency, and in part because Apple automated away many of the tasks done via GUIs in other environments.

Having the OS recognise and configure devices or services, for example, removed huge swathes of opportunities for user error and correspondingly reduced the burden on the GUI. Bottom line: as a GUI designer you don't have to handle what's already done -and that's true whether your goal is to provide comprehensive desktop services like Apple or command line developer support the way Sun does.

On the other hand, both Apple and Sun had all that figured out in 1984 and while they've gotten a lot better at implementing their ideas since, they haven't come up with a single fundamental improvement. In fact the only real candidate I can think of in that category is the lifestream project by Eric Freeman and David Gelernter at Yale.

Here's the summary from the project home page:

Lifestreams is a novel software architecture that was initially developed at Yale University. The goal of Lifestreams is to minimise the time users spend managing their documents and electronic events while increasing their ability to find and make use of this information. To accomplish this we have worked to create a software environment that parallels the way people work with electronic information and simplifies their electronic interactions. Lifestreams is built on a simple storage metaphor --- a time-ordered stream of documents combined with several powerful operators --- that replaces many conventional computer constructs (such as named files, directories, and explicit storage) and in the process provides a unified framework that subsumes many separate desktop applications to accomplish and handle personal communication, scheduling, and search and retrieval tasks. While our current prototype is tailored to managing personal information, a "lifestream" is also a natural framework for managing enterprise information and web sites; we are just beginning to explore such use.

Microsoft has been announcing adoption of some of the key ideas here since the late nineties and Apple actually has some core components, like contextual file search, working. Unfortunately a new desktop metaphor needs a consistent desktop OS to work with - perhaps a Pick or MUMPS revival?- because the overall effect of using it on existing products feels like reading French romantic poetry translated into English by a German -and to a cousin at that.

Notice, however, that I'm not saying that NeWs then, or MacOS X now, aren't better products, what I'm saying is that NeWs did graphics and networking better but didn't change how people interacted with the tools, that MacOS X really doesn't either, and that Vista is just more of the same ideas done differently.

I currently mostly use CDE (although KDE on Solaris is distinctly prettier and perhaps somewhat faster) while Sun, for reasons that baffle me, pushes Gnome, but none of these things really do much more for user productivity than SunView did - and the growth from something that ran nicely in 4MB with a 20Mhz processor to Gnome's minimum 160MB has to count as some kind of horrible record.

You could argue, for example, that GUIs are fundamentally about making it easier for users to choose functions and sub-functions while concurrently enabling multiple interactions on one physical screen. In that sense, however, SunView wasn't any less effective than KDE or Gnome are today. On the contrary, it was cleaner and simpler while nothing has yet beaten Sunview's great innovation: the toolkit driven walking menu. On my 19" Sun 160, for example, clicking an icon or the right mouse button could yield up to 12 choices in the first column, ten in the second, and eight in the third for a total of 960 choices without falling off the bottom of the screen - an order of magnitude more than the number of icons you can crowd onto a PC screen, and significantly better even than MacOS X does on the 17 inch Powerbook.


Paul Murphy wrote and published The Unix Guide to Defenestration. Murphy is a 25-year veteran of the I.T. consulting industry, specialising in Unix and Unix-related management issues.