I like the following best:

• No cascading menus so you lose the “misclick of death” feature that plagues cascading menus. Which is to say, when you misclick or move the mouse too far and you have 5 nested menus open, you have to start again. The SUSE menu changes what is shown when you open nested menus.
• When you are in a nested menu, a big back button the whole height of the menu shows on the left. This uses Fitts’ law too, so you just throw the mouse to the left side of the screen to go back. This feels so nice.
• All menu items have large extended selection regions, so you can’t miss them when moving the mouse quickly. They feel wonderful, the targets are huge.
• Moving the mouse to the bottom left of the screen opens the menu without a click being required
• Search is quick and omnipresent
• Menu items have the name in black, and the description in gray underneath

Other little things just ooze polish, like the “user username at domain” text having the username and domain in bold. This is slick.

Well done to SUSE.

The best thing about Gmail is the conversations feature. Emails are sorted into conversations. These would be sequences of back and forth replies between you and another person, or a mailing list thread. This means your average mail folder is neater and more concise. Best is clicking a conversation, it shows all mail that has been involved, your replies included, all-in-one-page. Being all in the same page saves you so much clicking around, it’s great. Mail you have read is wrapped up with AJAX, clicking displays the contents. Stuff you don’t need every read is also wrapped up, like full from and to information, when wrapped up Gmail shows a small concise info box.

There are no folders, just labels. They work like folders but mail can have multiple labels. There is also the archive, which hides things from the inbox. Labels can be assigned to incoming mail with a filter. Filters can archive and assign a label, and this is the best way to handle mailing list traffic and emails from Amazon, Dabs, etc. as it keeps your inbox clean and ready for important mail.

Folders show snippets of the conversation inline, to remind you of the content, or give you an idea about a new mail.

The number next to labels and the inbox show the number of unread or partially unread conversations, not the number of unread mails. This is so great for mailing lists as you now know when new threads start, which is the more interesting kind of mail when it comes to lists.

Spam filtering is great, better than Thunderbird. Presumably because Google learns about new spam daily since they have all the mail on their servers.

The famous contextual ads are often hilarious and well worth clicking. But easily ignored for those who are less tolerant.

You don’t need to click refresh, new mail shows up on your screen as Google receives it.

The title text reflects the unread inbox count, so you can just glance at the browser-tab to see if anything new has arrived.

Replies are inline, so you can read the conversation you are writing about as you type the reply.

As you compose mail, a draft is saved to Google’s servers every few words. Does Outlook do this? No? And even if it did you couldn’t stop writing at home and continue writing when you get to work without having to even think about clicking the save button. Not only that but drafts are saved attached to the conversation you started them at. So next time you visit the conversation, in list view or expanded view, you see “Draft” written in attention-grabbing red text.

Because all the work happens on another computer far away, the interface is frankly more responsive than any other mail client I have ever used. I think the future is totally thin-client and browser-based.

Google show great attention to detail, removing what limitations a browser based application would normally experience with clever solutions. Eg selecting all mail in a folder. You click the select all button, it selects the 50 conversations on the page. Then a small text is shown asking if you want to select everything else in the folder too. In a rich client you’d have all conversations in a big widget and would have selected them all by default. Google’s solution is an extra click, but you can still do want you want. There is a toolbar, and it is appropriately concise, but it scrolls off the top of the screen. Google have it replicated at the bottom too. Perfect.

In fact the limitations of the browser medium mean Google have innovated all over the interface. No menu system means they have had to be frugal with the interface, consolidating wherever possible and taking care with positioning. And they have succeeded. There is no menu for selection, but you have a row of links to allow selecting messages based on more criteria than most browsers allow. And yet nothing feels cluttered or unnecessary. I take my hat off to the UI team.

Apologies for the review being an large lump of text – I may improve it when I have motivation.

Appendix

Until then I was struggling with Thunderbird for personal mail at home on my linux box. Now it seems to me XUL products run like shit on Linux. *Soooooo Sloooooow* I blame GTK and Mozilla crapness. It’s annoying because XUL is so quick on Windows and OSX. Get your game together GTK for fuck’s sake, I’m fed up waiting for you to catch up with Qt’s speeds.

Anyway. It was annoying not having access to that email at work and at home, and consequently I was very behind with my personal mail. I tried to set up FreeNX, but gave up, despite having set it up successfully on another ArchLinux box at work. I needed a decent webmail solution.

Now obviously people mean the difference between, for instance, KDE and Gnome apps, and they mostly also refer to even more horrible toolkits like whatever it is that xine-ui uses.

But if you actually looked closely at Windows applications, you’d notice that every major MS application uses different toolkits and “skins”, and so do all the other applications from other companies. In fact Win32 only offers a few basic controls, namely:

• push-buttons
• scrollbars
• really basic menus
• combo boxes

The menus that win32 offers are so basic that nobody uses them, they implement their own so they can add things like icons, and menu-separators.

In my opinion, the reasons that Windows applications are consistent are:

• Consistent fonts
• Consistent colours.

Absolutely the number one reason is fonts. Windows offers a default font that GUI controls like menus, buttons and text widgets use. Apparently, the API for drawing text is different on Windows compared to X11 in that it developers end up using the right font when they roll their own widgets. On Windows, all text looks consistent.

Colours are almost as important. On Windows the colorscheme is a globally accesible property, on Linux schemes are toolkit/desktop specific. If toolkits just used the same colours, people wouldn’t notice small, or even larger differences in the widget styling.

So in conclusion, if Linux just standardised fonts and colors I think applications would feel consistent, just like they do on Windows. If toolkits did just these two things, their other differences would not be nearly as noticeable in general use. Frankly I feel if Linux could do these things, we’d be more consistent, since if you use mostly KDE apps, they have far more internal consistency than the Microsoft port-folio.

Disclaimers
Of course there are other more subtle considerations, like one toolkit on linux has 4 pixel bevels. It just looks ugly, there’s no way to hide that. But if it had consistent fonts and colouring it’d look a lot better and would fit in with a KDE desktop more easily.

It is also true that Windows appliations copy whatever is the latest MS-app appearance. MS Office 2003 has a neat style that has rapidly been copied and emulated in all other major toolkits, ie check out the latest version of Delphi, it looks similar to Office 2003, but the similarity is subtely different, eg. menus are narrower and more compact. And this makes sense since MS didn’t release any API or DLL to allow you to use the Office 2003 style, so Borland implemented their own clone. Trolltech even claimed to considering emulating it when I attended their roadshow in Cambridge. My point in this paragraph is that Windows apps emulate a general “Windows” consensus that is defined by MS, so this helps somewhat.

And yes, I’m just ranting an opinion here, maybe someone who reads this would like to do something about it, as I’m not the man for that job.

The game chair points out that this is exactly what happens in video games, with alarming regularity.

Highlighted is the Prey demo, a great new game that builds up a thrilling story, but at crucial moments, a new level has to be loaded and your immersion in the story is shattered.

A much worse example I think is HL2 and in fact any game based on Valve’s Source engine. Source games use an “immersive world” where there are no normal levels, just one super huge one. But you still get load screens, and in fact you get them every few minutes, so it’s much worse. There you are cruising along in your sand-buggy elated at the speed and the fire fight you just won and suddenly the game pauses while it loads the next section.

In Valve’s defence they planned the load moments appropriately so as to minimse disruption, but it still sucks.

World of Warcraft only has a load screen very rarely, instead it loads all the content just before you get there. Unfortunately WoW doesn’t have an immersive story/scenario like Prey or HL2 do, so it doesn’t really benefit in that aspect.

This seems fair and usable to me.

With regards to the politics of Google’s complaint, I must say I’m surprised how anti-Google the online world seems to be about this issue. Even Ars-technica who usually write what I’m thinking, write:

Sorry Google, this complaint rings hollow.

But I agree with Google, it is anti-competitive. Yes, Firefox and Opera have Google as the default in their search bar, but this doesn’t make it hypocritical for Google to begrudge the Internet Explorer default because:

1. Google don’t make their own browser
2. Google don’t ship their own browser as the default browser with their monopolistic* operating system.
3. Google could be displaced as the default search engine in Firefox and Opera.

Point 3 is one I’ve seen nobody else make, Google are the default in Opera because they paid for the priviledge. They are the default in Firefox because they sponsor Firefox developers and provide product placement/free advertising. They are the default in Konqi because Konqi developers picked what they considered the best search service. MSN-search could displace Google in all these products. But how could Google ever be the default in IE? They can’t because Microsoft are using their dominance in the browser market to promote their search engine, and there’s no business or technological method Google could use to be considered instead.

However you want to look at it, stuff like this is anti-consumer/anti-competitive/anti-you-and-me. Companies like Microsoft need to be legally restricted in what they can do, or other companies, even search-giants like Google, have too hard a job to step in and make the computing industry more interesting. And situations like this one make it too easy for MS to erode other search-engine’s market share without actually having to produce anything superior or innovative. What percentage of IE users will change the search engine default if they don’t actually have to make an explicit choice? Very few IMO, and that’s what scares Google and Yahoo! too.

Don’t you agree?

* Microsoft fit the wikipedia definition of Monopoly even if they don’t have a 100% market share and even though people can choose not to use Windows.

Codeine

Because I feel I should say something about KDE, I almost have Codeine 1.0.1 ready. I’ve fixed all the bugs I’ve been told about, and improved DVD support somewhat. Get your bug reports in today if you have anything else that needs fixing! I’ll release before the weekend.

Now I’m sure Office 12 is going to be impressive. I just hope they haven’t done quite so well with Vista, because although I’m (mostly) certain KDE 4 will be well researched, designed and implemented, I think it’s hard competing against the amount of money companies like MS and Apple can afford to put into design.

The blog is good. The author is obviously experienced and has led the UI development well. The entries are well written and very interesting! As someone interested predominantly in UI-design, I gained a lot of insight. I was a little thrilled to see ideas in there I’d been thinking on myself, which is also annoying as now people will just assume I copied if I ever implement any of them. And for sure KDE 4 should involve as many new widget-ideas and usage-concepts as possible IMHO.

Among other things he asks:

Why do modern Desktop Environments not exploit Fitts’ Law?

One of the implications of Fitts’ law is that there are five spots on a mouse-operated display that are easiest to target, the four screen-corners and the spot directly under the cursor. A default KDE setup with a maximised window has clickable widgets that respond to corner clicks. The K-Menu, the window-menu, the window-close button and the clock. Windows is the same. GNOME is debatably even better as all four corners always have something useful in them. But we could still make more use of the corners, they are after all extremely easy to hit. So why don’t the desktops have behavior associated with simply moving the mouse into the screen corners?

In my opinion, it’s because they are the easiest spots to hit with the mouse.

Setup your OSX box to trigger Expose when you move the mouse to a corner. Now count how many times during the day you nudge the mouse into the corner and trigger Expose by accident. It’s very useful to be able to easily trigger Expose, but it’s very annoying to be able to easily trigger Expose. The annoyance is such that I would definitely argue against inflicting such behavior on people by default. Apple apparently agree with me, by default you can only activate Expose with the F12 key.

Being Less Annoying

Launching a Konqueror instance, or KMail or doing something drastic like Expose would be annoying if done by accident. The new Windows would probably get in the way of what you were doing and Expose loses window focus and totally changes the screen contents. We want something that doesn’t interfere with what the user was doing, but still is useful enough that we are making more use of the corners.

How about non-intrusive, inactive windows that appear from the screen corners? They could show a calendar, or the weather for this afternoon, or your CPU temperature. Well, maybe, but we can’t show this passive-window in the same corner, or it would get in the way of the close button, or the K-Menu. So perhaps show the passive window in the horizontally opposite corner? Yeah I think that would perhaps be quite fun.

But this is hardly a revolutionary feature. The corners are the easiest spots to target on the desktop – you can hit them without looking at the screen! Yet, so far, my only suggestion that is both useful and not annoying belongs in the KToys module.

Better Uses For Hot-Corners

It seems to me we should trigger a really frequent task when the mouse hovers over a “hot-corner”. For instance: “Copy text to clipboard”. With some visual feedback, (perhaps Klipper shows a little non-intrusive popup that confirms the copy has occurred), this could be a nice feature.

But then again would the high rate of accidental activation just make it irritating? I get irritated already when I lose my middle-click clipboard contents because I accidently selected some other text. On the other hand I continue to use middle-click paste because it is so convenient despite its drawbacks. And even though the screen corners are mouse magnets, you don’t often have text selected, so triggering the feature accidentally would be less common. If the Klipper notification had a button in it labeled “Revert to ‘((previous content))’”, perhaps we would make accidental activation less annoying?

Perhaps I’ve got you thinking that I’m on to something. But generally when you want to copy some text, you’ve been typing, and having to move your hands from the keyboard to the mouse is bad ergonomics. But my idea would still work for copying a chunk of text from a web-page to the google-search-bar, etc. Perhaps the same corner could trigger paste if no text is selected.

Lets say we made the bottom right corner do this. We could even put the klipper applet in that corner to make the association more logical. But now imagine moving the mouse towards that corner with the intention of right clicking whatever applet is in the Kicker there. Can we imagine yourself getting more and more nervous as you realise you are getting closer and closer to the “hot-corner”? This is what makes me unhappy with the entire concept of just moving the mouse into a corner to trigger some function. I don’t think it’s a good idea to make users feel restricted about how they move the mouse!

But still I’d like to try out this copy/paste idea. Select text, hit corner, click paste location, hit corner. Quick and easy! Perhaps with practice I’d even learn not to accidentally trigger it. But I’m not convinced that hot-corners are a good feature for Desktops. I’m not surprised that hot-corners are not a default feature of most Desktop Environments.

Scrollbars

Changing topic slightly, there is one very important area KDE should take note of Fitts’ law. Right hand edge scrollbars. OSX and Firefox do it properly – if the window is maximised or aligned with the right hand screen edge, you should be able to use the scrollbar with the mouse from the right hand screen edge. The problem for KDE is Qt. QScrollView puts the scrollbar inside a QFrame with a 3-4px margin, so there is always a slight margin between the screen edge and the QScrollBar widget. You can set the QScrollView to be frameless and remove the margin, and this is something I would like to see happen with most applications. The problem is that the frame is often an important visual separator. Certainly with amaroK I think our interface would look a lot more cluttered without the frame. Maybe we need a three-sided frame? Or just a divider top and bottom.

Times Expose accidently triggered while writing this post: 3
And twice more during subsequent edits!