Blue95: a desktop for your childhood home's computer room

Computing is largely a cargo cult thing these days.

The problem is that the interfaces these bootleg skins draw "inspiration" from were designed on the back of millions of pre-inflationary dollars' R&D from only the best at Golden-Age IBM, Microsoft, Apple, etc.. BeOS, OS/2, Windows 95-2000 do not look the way they do because it looks good, they look the way they do because it works good, countless man hours went into ensuring that. Simply designing an interface that looks similar is not going to bring back the engineering prowess of those Old Masters.

I’m less inclined to attribute it to “these days”, as I remember the contemporary copycat themes in e.g. KDE and Tk looking off as well. Even Swing with the native look-and-feel didn’t quite look or feel right, IIRC.

As a (weak) counterpoint to supplicating ourselves to the old UI masters, I submit Raymond Chen’s observations from 2004[1] that the flat/3D/flat cycle is largely fashion, e.g. how the toolbars in Office 97 (and subsequent “coolbars”) had buttons that did not look like buttons until you hovered over them, in defiance of the Windows 95 UI standard. (Despite Chen’s characteristic confident tone, he doesn’t at all acknowledge the influence of the limited palettes of baseline graphics adapters on the pre-Win95 “flat” origins of that cycle.)

Also worth noting are the scathing critiques of some Windows 95 designs[2,3] in the Interface Hall of Shame (2000). I don’t necessarily agree with all of them (having spent the earlier part of my childhood with Norton Commander, the separate folder/file selectors in Windows 3.x felt contrived to me even at the time) but it helps clear up some of the fog of “it has always been this way” and remember some things that fit badly at first and never felt quite right (e.g. the faux clipboard in file management). And yes, it didn’t fail to mention the Office 97 UI, either[4,5]. (Did you realize Access, VB, Word, and IE used something like three or four different forks of the same UI toolkit, “Forms3”, among them—a toolkit that looked mostly native but was in fact unavailable outside of Microsoft?..)

None of that is meant to disagree with the point that submitting to the idea of UI as branding is where it all went wrong. (I’ll never get tired of mentioning that the futuristic UI of the in-game computers of the original Deus Ex, from 2000, supported not only Tab to go between controls and Enter and Esc to submit and dismiss, but also Alt accelerators, complete with underlined letters in the labels.)

[1] https://devblogs.microsoft.com/oldnewthing/20040728-00/?p=38...

[2] http://hallofshame.gp.co.at/file95.htm

[3] http://hallofshame.gp.co.at/explore.htm

[4] http://hallofshame.gp.co.at/visual.html#VISUAL36

[5] http://hallofshame.gp.co.at/visual.html#VISUAL38

> Despite Chen’s characteristic confident tone, he doesn’t at all acknowledge the influence of the limited palettes of baseline graphics adapters on the pre-Win95 “flat” origins of that cycle.

It's right in the second sentence: "...Windows 1.0, which looked very flat because... color depth was practically non-existent."

>they look the way they do because it works good

In modern times telemetry can show how well new designs work. The industry never forgot how to measure and do user research for ui changes. We've only gotten better at it.

Just because they're measuring doesn't mean they're measuring the same things as before.

The goal in 1995 might be "The user can launch the text editor, add three lines to a file, and save it from a fresh booted desktop within 2 minutes".

The goal in 2015 might be "we can get them from a bare desktop to signing up for a value-add service within 2 minutes"

I'd actually be interested if there's a lot of "regression testing" for usability-- if they re-run old tests on new user cohorts or if they assume "we solved XYZ UI problem in 1999" and don't revisit it in spite of changes around the problem.

With so many things being ad-funded, I always wonder if what they're optimizing for is "it took the user 50% longer to complete a task"

That assumes that they are using the telemetry to create a better product for the user rather than the developer.

Telemetry may tell you the "what" but, at best, it will only allow you to infer the "why". It may provide insights into how people do things, yet it will say nothing about how they feel about it. Most of all, telemetry will only answer the questions it is designed to answer. The only surprises will be in the answers (sometimes). There is no opportunity to be surprised by how the end user responds.

I've had an alternate theory for a while. Prior to verbose metrics, UIs could only be designed by experts and via small samples of feedback sessions. And UIs used to be much, much better. I suspect two things have happened:

- With a full set of metrics, we're now designing toward the bottom half of the bell curve, ie, towards the users who struggle the most. Rather than building UIs which are very good, but must be learned, we're now building UIs which must suit the weakest users. This might seem like a good thing, but it's really not. It's a race to the bottom, and robs those novice users from ever having the chance of becoming experts.

- Worse, because UIs must always serve the interests of the bottom of the bell curve, this actually is why we have constant UI churn. What's worse than a bad UI? 1,000 bad UIs which each change every 1-6 months. No one can really learn the UIs if they're always churning, and the metrics and the novice users falsely encourage teams to constantly churn their UIs.

I strongly believe that you'd see better UIs either with far fewer metrics, or with products that have smaller, expert-level user bases.

There's a much simpler explanation. At some point, the UI becomes about as good as it can be. It can't really be improved any further without changing the whole paradigm, and just needs to be maintained.

But product managers inside the large corporations can't get promoted for merely maintaining the status quo. So they push for "reimagining" projects, like Google's "Material Screw You" UI.

And we get a constant treadmill of UI updates that don't really make anything better.

Fuck telemetry. Don't spy on me while telling me it's in my best interest. Don't spy on me at all.

I don’t believe either is the primary driver of modern UI design. Cynical as it may be, I think the only things that get any level of thought are:

1. Which design is most effective at steering the most users to the most lucrative actions

2. What looks good in screenshots, presentations, and marketing

The rest is tertiary or an afterthought at best. Lots of modern UI is actually pretty awful for those mentioned bottom of the bell curve users and not much better for anybody else in terms of being easy to use or serving the user’s needs.

Proper use of analytics might be of assistance here, but those are also primarily used to figure out the most profitable usage patterns, not what makes a program more pleasant or to easy to use. They’re also often twisted or misused to justify whatever course of action the PM in question wants to take, which is often to degrade the user experience in some way.

> I’ll never get tired of mentioning that the futuristic UI of the in-game computers of the original Deus Ex, from 2000, supported not only Tab to go between controls and Enter and Esc to submit and dismiss, but also Alt accelerators, complete with underlined letters in the labels

I think that's because they used the stock UI toolkit of the original Unreal Engine, which also had all these things. If you recall, UT'99 actually had a UI more like a desktop app at the time, complete with a menu bar and tabbed dialogs:

http://hw-museum.cz/data/article/VGA-Benchmarks/Benchmark-VG...