Someone has tweeted this and got several retweets:

What they mean is this: when the content is still, tapping the screen is interpreted as a tap, but when the content is in motion, tapping the screen just stops the motion. So is behaviour modal? No, here’s why.

Most people think that an interface is modal when it has modes, i.e. when same user input produces different output depending on the state of the interface. However, that’s not the definition.

Let’s read Jef Raskin carefully:

An human-machine interface is modal with respect to a given gesture when (1) the current state of the interface is not the user’s locus of attention and (2) the interface will execute one among several different responses to the gesture, depending on the system’s current state.

Most people’s understanding includes only the (2), but not the (1). But they both equally matter. Perhaps, Raskin didn’t name the thing well, but we have what we have.

You unlock your iPhone’s and tap Messages:

But just as you are tapping it, you notice that it’s actually Shazam:

Oops, you are on a wrong page of your home screen.

In this case, launching of Shazam instead of Messages is a mode error: your gesture (tap in the top left corner) produced the wrong output depending on the current state (the page number), which was not your locus of attention. So, the iPhone’s home screen is modal.

Now let’s say you are in Contacts and tap the bottom left corner for Favourites:

Is there any chance you actually meant to go to a previously visited web page?

The gesture is the same (tap in the bottom left corner), and it produces different outputs depending on the current state (the active app). But here, the app is your locus of attention: you are fully aware whether you are looking for a contact or browsing the web. That’s why a modal error is not possible here, and this interface is not modal.

If we get back to iOS scrolling, it now becomes clear that it is not modal. When the scrolling animation is playing, it is the user’s locus of attention. The user is fully aware of the interface’s state: they are looking at the moving content. So the fact that the tap is interpreted differently during this animation is not a surprise and doesn’t produce a mode error.

It’s a common mistake to denote heading level by its font size only. The popular Bootstrap CSS framework does exactly that:

When you see this picture, it makes sense: the higher the level, the bigger the heading. The problem: this works only when you put these headings next to each other. But when you use six levels of headings in a real text, these styles will not help establish the structure. From the reader’s perspective, they would be just randomly-sized headings. To see an example, open the Bootstrap’s own documentation in the middle of its page.

Now look at this example from Wireless DJ help page:

The progression of sizes is not the only design tool here.

There is only one main heading (aka h1) on a well-edited page, and it’s on top. So from the structural point of view it doesn’t matter what it looks like. It makes sense to make it large though.

The second-level headings (h2s) are larger than the body text, but what’s even more important, they have generous margins around them, and the top margins are visibly larger than the bottom ones. This helps link the headings to their text and separate the sections from each other.

The third-level headings (h3s) are not just smaller than the second-level ones. They are actually of the same size as the body text, but are bold. Also, they have no bottom margins at all. This connects them very tightly with their text.

If you want more than three levels, you won’t be able to make do with just font size and margins. You will need to employ other tools:

This example is in Russian, but it’s even better if you can’t read it. Just see how headings of different levels have several attributes that help you recognise them: alternating typefaces, italics, rulers, numbering.

Such a complex hierarchy may be necessary in some legal documents (the example above is from the Russian Civil Code), but in most cases, three levels of headings must be enough.

All screenshots for this post were made by Rakhim Davletkaliev.

The main interface

The canonical view of two directories in the Left and Right panels:

The right panel is active — you see the cursor and the highlighted path in the title. Use arrows to navigate the files. Change currently active panel with Tab.

Alt+F1 or Alt+F2 changes the displayed volume in Left or Right panel respectively:

MS-DOS uses drive letters for volumes, and the letters A and B are reserved for the floppy drives. Most computers had just one, and then the main hard drive was C.

Hide the panels with Ctrl+F1 or Ctrl+F2:

In MS-DOS, the filename’s length is limited at 8 characters, and the extension, at 3. The dot between the file’s name and extension is not shown. Unlike the command prompt, Norton Commander displays filenames in lowercase letters and directories in all-caps. There is a special halftone pattern separating the extension of the system files like Io.sys and Msdos.sys (also they have their first letter capitalised).

Ctrl+O hides both panels, for work with the MS-DOS prompt:

The main menu

The main menu is the one displayed on the bottom, with keys F1 to F10 mapped to the most-used functions.

F1 for Help was a standard for many MS-DOS programs. I’m not sure if Norton Commander originated it.

F2, User Menu. The items can be programs for quick access. Here, the User Menu is empty:

The user has to edit a text-based configuration file to add items to this menu (press F4 while in this F2 menu):

The main panels are blue, the menus are cyan and the dialogs are grey. Unless they are error messages:

F3, View. Opens text view of the currently selected file:

Of course, for executables it shows “garbage”. But as far as I remember this version has the ability to set up external file viewers, so you can make it display images as actual images.

F4, Edit:

See the Mark item in the main menu inside the editor? It has something to do with a clipboard. There was no system-wide clipboard, of course, and the programs that had this feature used very different keyboard shortcuts for it. In Turbo Pascal 5.5, I remember, there were crazy combinations like Ctrl+K,B and Ctrl+K,K.

F5, Copy:

The Copy command opens the Copy dialog where the user enters the destination path. The input is pre-populated with the name of the directory in the opposite panel, and normally the user doesn’t edit it.

F6, Rename or Move (shortened to RenMov in the menu line):

Why Rename or Move is one command? Because from the system point of view it’s the same: while you are within the same drive, it just changes the full name (i.e. including the path) of the file. Again, the input is pre-populated with the path of the directory in the opposite panel, so just press Enter to move there. Or type a name instead of the path to rename the file in-place. Or add the new name to the end of the path to move and change name at the same time.

Moving between volumes involves actual copying and then deleting the original, but it was also done with this menu.

F7, Make Directory:

Interestingly, the window has no confirmation button, unlike the previous ones. Just press Enter. Also notice that in the main menu bar it’s called Mkdir, not MkDir. Why? Probably, because that was the name of the MS-DOS system command to make a directory.

F8, Delete. No question mark:

Before invoking the file operations like Copy or Delete, you could select many files with the Insert button, and they would display in yellow:

When files are selected, the bottom line changes to reflect this state.

F10, Quit:

The pull-down menu

The top menu, or the “Pull-down menu” is shown when the user presses F9 (or clicks the top line of the UI, if your computer has a mouse).

The Left menu sets up the Left panel. You may see many options of what to display in the panels, then how to sort it and more:

The Files menu lists many elements that have nothing to do with files, including Quit — not dissimilar to how the File menu is used in today’s applications:

Disk:

Commands:

Right, same as Left:

Panels setup

As you see in the menu, the panels could display a Brief or a Full view. The Brief view just shows the files, in three columns (as in the previous screenshot). The Full view shows one column of files with details:

In this mode, each line looks exactly like the bottom line in the panel, so why not just remove it to show two more files?

You could also set up file filters:

Use a panel for file preview:

Or for search results:

Or for item information. Directory:

File:

Go to file (in the current directory) by typing a couple of letters of its name:

Other dialogs

Compress:

Find File. The dialog windows that have other dialog windows inside, are cyan, not grey — for contrast:

The inputs are shown with square brackets and dots between them. The focus is shown with a black background. Advanced search options — this window is grey:

Notice how checkboxes differ from radio buttons.

History:

NCD (Norton Change Directory? I’m not sure), a fancy way to navigate the folders on your disk:

Synchronise directories:

Format diskette:

Copy diskette:

Label disk:

Commander Link:

Commander Disk Cleanup helps find files to delete:

Honestly, I don’t remember this feature at all (but I didn’t use Norton Commander at that time because there was DOS Navigator, which is a whole different story).

Found something. Here, files are displayed with the dots in the names:

System information

System information:

Memory summary:

DOS memory blocks:

CMOS values:

Configuration

Now let’s look into the configuration dialog. Notice how the controls are organised and grouped:

By the fifth version, Norton Commander got colour schemes:

The palette was limited to 16 colours. There was no way to make the colours darker in the shadows, so to simulate the darkened cyan-on-blue or white-on-cyan the applications of the era used grey-on-black. I’m not sure, but I guess Norton Commander has borrowed this visual style from Turbo Vision. The earlier versions had no shadows. If you go back to the Main menu section, you’ll notice that the help window has no shadow for some reason.

EGA Lines is a mode where twice as many lines of text fits the screen. Nobody I knew ever used it:
.

Fun fact: this post has 57 full-screen full-quality screenshots. Their aggregate size is 271 kilobytes.

Building numbers sometimes double as street lights, as they do in Helsinki:

1

Sings with arrows help find buildings inside the blocks:

2

3

The path on the right leads to the amazing Berghain nightclub:

4

Non-standard street name plate and a plan of the quarter:

5

In-quarter wayfinding with colourful shapes:

6

Street name is put on a wall. Schwedter Straße:

7

Bernauer Straße:

8

There’s a lot of bike rental shops with handmade signs like this one (there is still an “official” city-wide bike rentral):

9

“Window-blinds like these are prohibited”:

10

More signs and plates:

11

12

13

14

15

16

Apparently, a burger place is coming here:

17

Beautiful:

18

Nice posters:

19

Crazy posters:

20

Konzerthaus Berlin. Look at the event posters:

21

22

Photos are from the trip of May 2016.

More Berlin:

• Street name plates
• U-bahn
• U-bahn station names
• Coffee
• Planet Modulor

See an earlier post on Berlin U-bahn station lettering. Now, to the rest of the U-bahn.

Typical station entrance:

There is also a tram stop to the left and a train station above the road.

Stadtmitte station has its own entrance sign design:

A schedule at the entrance:

Elements of wayfinding inside the stations. The list of the remaining stations of a line:

Another one:

No turnstiles, as usual in Germany. To the left of the ticket machine (white and yellow), there is a small validator (yellow and white):

At the centre of a platform there is usually an element with the station name and exit guide:

A similarly design element, but wall-mounted:

Another exit guide design:

Time to the next train, service information and a clock:

A compact version:

Ads:

Train are yellow:

All trains have an external speaker (above a window) so that the announcement are heard from the platform. The windows are filled with the pattern of Brandenburg Gate.

Inside an older train:

Inside an newer train:

Line diagrams are grouped. 5, 8 & 9:

6 & 7:

Apparently, this reduces the production cost. But this format also makes the shown lines more “related” in the eyes of a passenger, even though transfers to other lines are no worse.

A button to open the doors:

The system map is on the ceiling:

Hermannplatz station:

Weberwiese station:

A train on an elevated structure:

The photos are from the trips of March and May 2016.

More Berlin:

• U-bahn station names
• Street name plates
• Coffee
• Planet Modulor

These are words that usually signal problems with the user interface.

Main and secondary

Variants: regular, general, basic, advanced, extended, miscellaneous, more
Examples: main section, basic options, advanced plan, miscellaneous settings

The line between “primary” and “secondary” is not defined. The user is looking for a particular thing and has no idea whether it’s “main” or “advanced”. “More” is usually a graveyard of elements that the designer didn’t find place for.

Useful

Variants: important, notable
Examples: useful hints, important information, notable changes

When you nominate some stuff important, this means all the rest is unimportant. Instead of stating the usefulness, explain the benefit: “How to start snowboarding”.

Article

Variants: post, entry, publication
Examples: add entry, next post

These, again, name the type of the content. The words can be useful among the editorial staff, but meaningless for the reader: there are no newspapers with an “articles” section.

Catalogue

Variants: list, archive
Examples: shoes catalogue, news archive

There’s no need to signal that a list is following. Just put the list with an informative heading: “Shoes”, “News”.

Click here

Variants: enter, select, go, follow, open, launch
Examples: click here to open, enter password, select country, follow the link

There is no need to explain how to use buttons, links, input fields and other standard user interface elements. Links should just name the places they lead to: “iPhone 7 review”. Form fields should just name the content: “Password”, “Country”.

Form

Examples: application form, inquiry form

A form is a table of fields to fill. The word “form” just names a type of screen, but the user already sees that it’s a form. Name it with the benefit in mind: “Job application: designer”.

Required

Variants: necessary, must, please
Examples: required field, you must agree, please specify the phone number

The user doesn’t care if a field is required. If the form doesn’t work without it, they will put in some garbage. Instead of demanding or begging, explain the benefit: “We will call to coordinate delivery time”.

Operation

Variants: process, transaction, request, step, state, module, function, data
Examples: process is not responding, bad request, step 5 of 12, module not installed, wrong data format

These words are handy to describe how something works technically. But there is no point in using them in the user interface: they just complicate the matters for the user. Write as a human being: “Spell check available in paid version”, “Due to an error, the app needs to re-open”.

Authorisation

Variants: authentication, authentification, identification, session, limit
Examples: please authorise, session timeout

Even programmers mix up the autho-whatever-s constantly. Use the verbs “to enter” or “to sign in”, or, even better, name the thing that’s inside: “Shopping history”.

Success

Example: operation completed successfully

If operation hasn’t completed successfully, it hasn’t completed, period. Write what’s been done: “Money sent”, “Update installed”.

If you know why a particular word from this list is not good, but in your case it makes perfect sense, leave it.

I visited three coffee places in Berlin: The Barn and two Bonanzas.

This is The Barn (Auguststraße 58):

Espresso, flat white and a cappuccino (I had the flat white as I always do):

Sugar (not for me):

This is the Bonanza at Mauerpark (Oderberger Straße 35). Great sign:

Inside:

Flat white and water:

This is the Bonanza in the yards of Friedrichshain (Adalbertstraße 70):

Nice building and the tables outside. Here’s their central office and the roastery.

Inside (menu on the mirror):

Almond thing, water, flat white:

Tasty pastry in coffee shops always tries to distract you from the coffee. I try to resist it, but don’t always succeed.

And this is not coffee, but a wrap and a smoothie in a vegan place Goodies (Warschauer Str. 69):

The photos are from the trip of May 2016.

More Berlin:

• Street name plates
• Planet Modulor
• U-bahn station names

A slider is a simple user interface control where you adjust some value by dragging a handle in a groove:

Most web developers can’t get it right. You may think that there is not much you can fail at with such a simple thing. But most sliders are bad. They don’t respect the Fitts’s law and don’t provide decent feedback.

I decided to write a manual. Let’s see what usually goes wrong and how to fix it. If you are reading this via RSS, please go to the browser to see the demos.

Fitts’s law

Common mistake is requiring to grab the handle to drag it. Logically, this makes sense. But a small handle is very hard to grab. Try here:

A tiny bit better is making the groove clickable:

But in this design the groove is so thin that it hardly adds anything — aiming is still a pain, particularly on a touch screen. Speaking of touch screens, some implementations would just forget about them and not handle the touch events.

You should be able to grab the handle from any point in the slider area:

Make this area at least as big as a comfortable button would be. Notice how the areas to the left and to the right of the slider also work, moving the slider to the minimal and maximal positions.

Feedback

In all the examples above the mouse cursor was changing to a pointing finger when you hovered over the draggable area of a slider. Sometimes this feedback is absent or inconsistent. Here, you can drag from anywhere, but the mouse cursor changes only over the handle:

Having no sign that the slider would work otherwise, many users would try to grab the handle. An even worse mistake would be requiring to drag the handle but displaying a pointing finger over the whole slider area.

Use feedback consistently, change the cursor over all slider area and make it all work:

The change of the cursor is the minimal feedback necessary for the user to understand how the slider works. But it’s better to also highlight the slider itself:

Notice how this one just feels nicer to use.

A small detail: the slider gets its red “hover” highlight, but also keeps it while being dragged, even when mouse is outside.

Continuous feedback

The slider should move continuously as I move the mouse. Some implementations would only repaint the slider when I release the mouse button:

Others would accept clicks in the whole slider area, but only move the handle to the click position instead of grabbing it (unless you grab the very handle):

Both feel broken.

The slider is usually there to control something else. In this case, the rectangle to the right of it. In some implementations, the slider would move continuously, but the external changes would happen only on release:

Everything should stay in sync: hover effects, the cursor shape, actual active areas; reaction to click and drag; feedback inside and outside the slider.

Feedback and slow data

Sometimes it may take noticeable time for the changes in slider to take effect somewhere else. You may need to make complex calculations or get data over the network. You may think you just can’t provide continuous feedback in such cases. But you must still aim to provide as much continuous feedback as possible.

In my examples, the slider controls the numeric value and the background colour. Let’s pretend they are slow to update.

In the worst implementations, the slider would get repainted only when the data is ready:

This is painful to use.

At least the slider itself should repaint continuously during the drag, no matter how long it takes for the change to take effect:

But we can do better. To fake continuous zooming, Google Maps shows at least a blurry map image (see my post on immediate feedback when data is unavailable).

What if only the colour is slow to update, but the number can change instantly?

Again, compare with the previous implementation — this one just feels snappier. So: always consider at least partial continuous updating. This is much better than waiting for a full update before making any change.

Notice that the data updates as fast as it can without the need to release the mouse button. If the data takes even more time to update, some progress indicator may be used, but the slider should never become unresponsive.

Reference implementation

There are too many sliders above. Which are the good ones?

This is the best, with full continuous feedback:

This is the second best, with partial continuous feedback when some data needs time to update (colour, in this example):

Please show this to your web development team.

In Sayve 1.3, we added the ability to display a badge with the number of unsorted audio recordings:

This is for those of us who don’t like their audio recordings to pile up. Actually, we wanted this in 1.0, but it didn’t get there. Why? It wasn’t that easy to design.

OK, this must be laughable: what even is there to design? The badge is a system feature that has a fixed look, just take it. But you can’t.

This is going to be a long and boring post, by the way.

Since iOS 10, displaying a badge requires an app to have asked user’s permission for sending notifications. So if we just tried to just show it, the user would see the system dialog asking for the permission.

We already ask for two permissions on first open: microphone and speech recognition. For iOS these are two separate things, unfortunately. Before asking, we show this screen:

This prepares the user for the system questions and helps us get a “Yes”. Microphone:

Speech recognition:

Notice that we can also add some of our own text in these dialogs to explain why we are asking for these permissions.

We could add a “tutorial” screen before asking about notifications: explain that we are going to display a badge and only then ask for permission. But adding another such screen and third permission question would make the first-launch experience really clumsy. And what’s more important, many users would still presumably press “Don’t Allow”, because at first launch they have no idea about how they would use Sayve and why a badge might be useful to them.

And what should we do if the user answers “Don’t Allow”? You can ask for a permission only once. How would the user even know they could change their mind later? For comparison, when the user answers “Don’t Allow” to any of the two sound-related questions, Sayve just cannot work, and we show this screen until both permissions are given:

We can’t do the same for the badge permission, because the badge is not necessary, it is just a preference. Some users would like it, others would not. So if the user says “Don’t Allow”, we would need to explain how to enable the badge later. But where and when? We don’t even have an About or Settings screen.

We could go and add a Settings screen just for this one setting. See how this is getting complicated?

Adding in-app Settings screen makes little sense: iOS already has notifications settings, and this is the only “notification” feature we use. So the next idea was to make this setting configurable only in the Settings app:

Let’s just display a badge when the user has enabled notifications. And we could explain this option in the app’s description.

This also didn’t work. Turns out, you can’t just put this “Notifications” element to the apps settings screen quietly — it would only be added there after your app has asked for a permission. So we have to ask for a permission even to get a “No”! By the way, the question starts with the text “Sayve Would Like to Send You Notifications”, which is not true:

We would not like to send any notifications, and we don’t even have them. As to the badge, we just can, but not would like to use it. To make things even worse, in this dialog, unlike the two previous ones, you cannot even put your explanatory text.

But this is not something we can change: the question is asked by iOS, not us.

So we added this icon to the top right corner:

The idea was: when the user taps this icon, we would first ask if the user would like to see the badge, and if yes, we’ll ask for the system permission to enable notifications. And we would remove the icon then.

Do you think this worked? Of course not. What would we do if the user declined the first time? There would no way to add the notification permission setting to the system Settings app then. So we cannot ask the user if they want a badge, we must just tell them that the notification permission dialog will follow and it is there to display a badge if they want.

So, in 1.3, when you tap the icon in the top right corner, you’ll see:

And after you press “OK” (which is misspelled as “Ok”, unfortunately), you see the actual, system notification permission dialog where you can decide whether you want the badge or not.

After having shipped this, we see that it’s better to say “Continue” instead of “OK” in the box above, otherwise it may seem to the user that they have no choice.

Is this the best design of a feature? I don’t think so. But that’s the best that we could come up with given the system limitations that we have.

We hope you like Sayve (we are Mihail Rubanov, the developer, and myself).

The button:

Get in the App Store