Let's start over

First of all, yes, button and entry mode are gone and won't come back in the near future.
Presenting my work to the broader audience last week resulted in the worst scenario I could ever think of. Last week I spared to effort to improve the GUI and navigation. Some ideas came from the feedback posted in my blog or in the bug tracker, too. Also, I was frustrated at first, the feedback really helped.

My work resulted in a completely new GUI. Furthermore, I improved navigation. You can now use Deskbar-Applet quickly without touching the mouse. If you typed in a search term you can jump to the search results by pressing the down key. Pressing enter while a match is selected will launch the default action (much like old Deskbar) and pressing the right key will show you a list of actions. A match has more than one action can if a '>' is on the right. Navigation with the mouse has changed, too. Clicking on a match will execute the default action (again much like old Deskbar) and pressing control while you click will show you a list of actions.

Nevertheless, I still have a problem with focusing the actions treeview. It works fine when I navigate with the keyboard. If I click on a match with ctrl pressed to show the actions I want to focus the TreeView that way that the first item is selected and I can navigate without clicking on the TreeView again. I thought grab_focus() would do it, but the first entry just stays grey and I still have to click in the TreeView. Any help is highly appreciated.
Last but not least I updated my guide Writing new-style modules for Deskbar-Applet.

Finally, please download and test the new release and tell me what you think about it (especially the GUI).



To be perfectly honest I don't like this new deskbar applet esp, and I know I'm not the only one; some of my friends I've talked to share the same view. The applet is way more complicated and more difficult to navigate around and didn't invade my entire screen. Since this new deskbar applet changes everything why not fork it and call it something else... since no longer is the deskbar applet many came to love.

In fact, don't even call it an applet... there's no need to have it be applet, since it doesn't actually do anything in the tray besides open a new window. So might as well bind the stand alone "deskbar replacement app to F3"

While I know the old one was slow and inefficient it was pretty intuitive. The new deskbar applet is simply useless to me.

-- Just another frustrated user.

In reply to by Milosz Tanski (not verified)

Okay, I see a lot of disgruntled and whining users here. So here's my two cents: It's _very_ impolite to whine about free software. You like it, you fix it, or you live with it. Deskbar is written in python the last time I checked, so no excuses.

Anyway. Out of pure curiosity: Why where the previous two modes removed, and: would it be okay with you and other deskbar developers if an outsider wrote the code to return them (for after 2.20)

In reply to by Jauco (not verified)

I absolutly agree with you. If you dislike what I did please help to improve it. I can't understand those people that claim for a fork before they even tried to improve Deskbar itsself.
The reason why we decided to support only one single new UI was that it was and still is enormously difficult to get navigation and focusing right. I don't want to go in detail here, but beleave me it's everything else but easy.
Of course, any help is appreciated. You can write your own GUI if you want to. In fact, it is a lot easier with current architecture than with the old one. If you really want that your ideas get reality step up and supply patches.

In reply to by sebp

We are not whining about your efforts.

We are whining about:

- the decision of gnome to include this at this very late moment, without the most important plugins ported to the new system.
- the fact that it replaced something that it has got nothing to do with. Its not a bar, its not on my desktop and its shouldn't be an applet either.

It's free software yes. So we have no right to whine?

Not exactly. We invest time learning tools. We choose to become dependend on them. We do this because we, the users, got the impression that our experience mattered. Also, even though its just a community, we were given the idea we get some sort of support. And for most part of the free software world, that's true. I wonder how many people would use free software if the official policy would be "GNOME: Don't like it? Fuck off!" .. I don't think many would.

This project isn't just somebody's hobby project and we can use it if we want to. It's part of GNOME and like gnome, should have the intention of at least trying to satisfying the expectations they set out to. Not misleading like making us dependent on a tool and then remove that tool to replace it with something completely different.

About the usability:

I'm using Gutsy with the latest updates and your new version isn't synchronized yet, it seems. However, with the previous version, these usability issues existed:

- it takes more that the original 1 click to select a history item (4 clicks, yuk)
- selecting an item doesn't close the result-list by default (this is expected behaviour given the purpose of the tool)
- resizing issues obviously. When enabling the history pane, the vertical scrollbar of the results moves so far right you couldn't see it no more.
- it doesn't close on [ESCAPE]
- the default action is assumed to be the first one, but it isn't highlighted or anything!


If any of these methods require fewer clicks _or_ fewer seconds deskbar will not be used to launch an item.

- pressing alt+f2 and typing something pressing [enter]. Deskbar should at least be just as quick with just running a command!

- using the menu, it takes 3 clicks to launch a program. Using the history should take 3 or LESS clicks.

- using tracker search, typing something .. launch document. Deskbar should be at least as quick to do the very same thing. Or it won't be used.

- when the window doesn't close by default, that's another click added to the score. Most users won't bother looking in the preferences to fix that. They just won't use it.

- It's not a problem that launches a complete window. But make sure to set the window-type so that compiz uses a quick animation. Try setting window type to MENU or POP-UP or something like that.

- It should be accepting my input before the window is shown. So i click and immediately start typing. Perhaps it takes 0.5 sec to show the window. Just make sure those keypresses are registered.

In the end, it could be just as usable as the old deskbar was. It's not about it being a window instead of a button or a bar. It's just important to focus on usuability rather than bells and whistles..

In reply to by Anonymous (not verified)

The new version takes away the majority of my issues with the new deskbar.
Sorry, if I and many other bug-reporters sounded so harsh!
We were just very worried and very used to using deskbar daily.

Some positive points to show you love:
1) it loads fast (perhaps its the lack of menu?)
2) it is completely keyboard controllable!
3) it goes away when you press escape!
4) history items are now only 3 clicks!

Still there is room for some small improvements:

1) alt+f3 doesn't hide it .. it should work like a hide/show thingie

2) when you press the right arrow-key and there are no alternative actions it just executes the action. Usually by accident. Only execute on [enter] please.

3) since the keyboard interface is fast and slick, make the mouse interface more 'explorable'. Add a 'clear entry field' button. Perhaps a preferences button?

4) when the list is empty it does not show the history, this would be a good idea. You can remove the history dropdown thingie in that case. Keep the clear-history button though.

5) when an action is executed you can clear the search-bar (like when we press escape)

6) there is no intuitive way to get to the alternative actions using the mouse. Perhaps that should just be the default action anyway? There is no way to get back with the mouse either.

7) we still launch stuff with an (accidental?) single click. Make it double-click. Double-clicks are not slower than single clicks, but they make the intention explicit. Keyboard is for fast usage, mouse for controlled usage. Single-click is just too much by accident.

8) search-plugins from firefox can be united. Default action would be the primary search engine, the secondary search engines available as alternative actions. Same with yahoo search results. These two trim down on the list a lot! Files and such should still be in the first list off course. Can the alternative actions for files, offer go-to-location?

9) perhaps a generic search category? Search for filenames, search on google, search with tracker? Instead of listing search plugins together with web history and such.

10) it seems to remember scroll position per query? Is this the expected behavor? It took quite a while to figure out why it sometimes scrolled a bit. A bit unintuitive.

I can't think of anything else. Yeah, there are still 2 bugs:

1) the libdeskbar-tracker module generates a warning from apport about deskbar crashing. Although it launches just fine. I guess this is to be expected with the tracker-plugin to be out-of-date.

2) when the window looses focus it does not hide. This means I have to click twice to see it again! Your previous version did hide! I assume this was the intented behavior.

3) sometimes (perhaps history related?) it stops showing results. Not even the default results it always shows. It doesn't happen when I clear my history and turn suggest history stuff off. Something wrong there?

4) when I remove the deskbar applet from the panel, it is still running and responding to alt+f3.

Thanks for all your hard work!

In reply to by Anonymous (not verified)

I knew all the changes were being done for a reason but I was saddened by the first windowed release non the less. I was shocked to the point where I(out of the loop) thought it must have been a rogue bug. I am glad an active dev is here to take these issues and work with them

After this release I am very happy to say the app has regained much of its usability. Sadly not having bar entry or the pop down function still sucks but the trade was worth it for now.

I would like to thank the developer for not taking all of our sadness and complaining personally, and seeing it for the productive critisim, by loving users, that it was. Also I am very happy with the improved code and the new head room you have added for the future. I hope many of the plugins are ported over to the new API before any of the major releases.

I have configured my window manager to force the deskbar to be decorationless and placed directly under the panel. Thank you compiz.

In reply to by Anonymous (not verified)

Thank you so much for all your ideas. I already added some of them to my todo list.
sometimes (perhaps history related?) it stops showing results. Not even the default results it always shows. It doesn't happen when I clear my history and turn suggest history stuff off. Something wrong there?
I didn't cam accross that problem, yet. I need more information to figure out what that might cause. Please report it to bugzilla.gnome.org then.

In reply to by sebp

"Please report it to bugzilla.gnome.org then."

I will try to replicate the issue first. If I find reproduceable steps I will post the bug here. Perhaps its just old history interfering since I can' t seem to reproduce it now. But if it still exists it will show up. I'm back to using deskbar daily ;-)

Some extra suggestions:

1. Drop the categories (or at least offer the posibility to do so)

Their original purpose was to cut down on the list of matches selectively. But your action-list already takes care of that in a much better way!

I've already suggested to put firefox-search-results and yahoo-live-results as ONE match, with the extra matches as alternative-actions.

To expend on this idea, please let _us_ select the primary search engine instead of using the one firefox is using. Firefox has no 'primary' search-engine, just a last-used search engine, which isn't the same thing. This will also make the 'sync' issues with firefox about which search engine is primary go away.

2. Drop the window borders, and maximize the window. Just fill the complete screen. Either history is showing, or search results are.

This will make it feel like button-mode again. It also has the same look'n'feel as the search button in nautilus. Lastly, it ends all discussion about resizing and what the perfect default width would be.

Unless I'm mistaken, not having a window chrome is just another hint like skip-taskbar. Maximizing by default shouldn't be an issue either.

In the future we can look at some transparency patch when compiz is enabled.

3. Perhaps we don't need a clear history button after all. It's already there in the right-click menu.We do need a clear search entry button. Empty search == Show history in the list. Also possible: just offer a clear history as an action item when showing the history! Make sure to confirm it though. "Do you really want to clear your history?"

4.When in the action list. There is no way to get back with just the mouse, nor is there feedback which match we selected. Some possibilities:

- a crumble-style-bar like Nautilus uses that replaces the search entry when we are in the action-list.
- "go back" as first action, although it should not have focus by default.the second item on the list would be a category named after the chosen match ("Search the web..") with the action as child-nodes.
- replacing the search-entry field with the chosen match and a back button.

Keep up the good work!

In reply to by Jauco (not verified)

It's called user feedback, and usually developers appreciate it. The "If you don't like it, fix it yourself" mantra usually only applies when someone is asking for a new feature or change in functionality, not when somebody is asking for existing code that used to work to be left in the program.

Also, given the comments so far from the maintainer, I get the impression that any contributions to bring the old-style back would not be merged. He can correct me if I'm wrong on that.

Even if I am wrong, it seems like it would be too late to get them in Gnome 2.20 and therefore Ubuntu 7.10. Basically this last minute addition has left no time for those "fix it ourselves" types to "fix it for others".

Is the decision to replace on-panel and button entry modes with a window deliberate or have the just been broken during development and we are supposed to wait for them come back? Simply, define "near future"

I really think that these modes were much more elegant than the new window (It might be just a matter of me being used to the old modes, though).

In reply to by sebp

Does that mean that patches to bring back the old interface will be rejected or that the project developers will not bring back the old interface?

My questions boils down to will we need to fork or patch?

In reply to by Anonymous (not verified)

You can always submit patches! If the patch looks good and other developers agree it will be commited, of course. I won't reject patches as a matter of principle.

I am sorely disappointed with the disappearance of the entry mode. This constitutes an exceptionally large UI change that I do not feel was properly considered.

This forces me to change how I work with deskbar for no appreciable benefit (yes I have tried the new versions, I am running from SVN)


Just to summarize some of the comments here: http://bugzilla.gnome.org/show_bug.cgi?id=465658 and here: https://bugs.launchpad.net/ubuntu/+source/deskbar-applet/+bug/131446.

Why change the UI radically one day before the feature freeze? It makes it kinda hard to make UI fixes and to ensure that your UI is usable by a LARGE number of testers. You admit there are problems, yet they are now frozen in.

The old UI was easier to use than Alt-F2 (the Run dialog). Now Alt-F2 is easier, and it will close with ESC.

Deskbar-applet was a utility, to make other tasks more accessible. Now it is Desk-App, an application that impedes the workflow it once accelerated.

I appreciate the work you did to refactor and improve the application, but making such an enormous change so close to release (effectively one day, due to UI Freeze) was an unfortunate move. I (and many others) want to continue to use the Deskbar, but can't in its current state. Unfortunately, I'm going to have to revert to using the contact-lookup-applet and the mini-commander applet.

All is not lost: http://live.gnome.org/ReleasePlanning/RequestingFreezeBreaks

Best of luck to you.

i'm sorry, but i have to agree with the others.
i've been using the button mode and hoped that it would still be available in the new version, although i haven't seen it on any of the new screens. now it's clear why...

i really hope button and entry modes will be back in follow up versions.

I do like the new interface, actually. But does it really need a panel button now that it is a standalone app? Could it not be accessible through a command that could be launched by pressing alt+F3 or even alt+F2, thus -sometime in the future- replacing the current run dialog?

Maybe keep the panel button, but make it optional. Put it there for folks who don't even know how to use alt+F2, but let the rest of us keep a clean panel.

I must be in the minority in that I do like the new UI, especially since keyboard nav is somewhat back in the latest versions.

I thought the goal of Deskbar was to be more than just an app. launcher(alt+f2) and the new UI seems to be a start in that direction(think quicksilver).

Just a thought, it would rock if deskbar could detect that it is being run in a Compiz environment and the UI would switch to "compiz enabled look" like the volume control does when access via keyboard shortcuts.

First of all, yes, button and entry mode are gone and won't come back in the near future

Does this mean that for those who want a utility rather than a huge quicksilver type application will need to move on? I fail to understand the logic in not spinning your new application into a new project rather than turning the deskbar utility into something like this. There was no reason to destroy something that worked very well.

So as I understand it, the only option now is to fork deskbar so the many people who would like it can have it back?

In reply to by Anonymous (not verified)

It only worked very well from a user's point of view. The code to get it working that well was a huge hack and very difficult to understand. That made the code very difficult to maintain. We went that way to make deskbar much more maintainable. In my opinion Deskbar will profit much more from that.
I really don't understand why some people think making a fork solves the problem! You know that you can submit patches and participate in developing in Deskbar?

In reply to by sebp

> It only worked very well from a user's point of view.

And that is the whole point. It worked great for users. So well that it is perhaps worthwhile to maintain the application.

After all, it is a _user interface_ application.

> The code to get it working that well was a huge hack and very difficult to understand.

Fine. So a refactor should make the code easier without losing the functionalily

Without. Losing. Functionality.

As has been said, please don't take the criticism personally - instead thank you for working on improving the deskbar engine and refactoring the code.

On the other hand, one of the major selling points of the deskbar applet has been the presence on the panel, as a text entry bar. You must realize that the decision to remove support for that undermines a great deal of the popularity the deskbar has enjoyed, even if you personally do not use it that way (together with many others, I'm sure).

The old entry interface made Deskbar one of my top-ten favorite apps in GNOME. Now, it's just not as interesting.

Please consider implementing the textbar support. Or, at, least, come up with a better name so that people won't think it's a deskbar, because right now it is not. Maybe "Command And Control"? :-)

Is it possible to remove the window border to the GUI, and link the bar to the panel applet? in this way it is really an applet. Now it doesn't have much to be an applet... it seems more an application!

Hi! I appreciate that you are working on deskbar. A popular project needs care and maintenance, and lots of bugfixing.

However, the UI changes are wrong. Don't ignore your users -- the main attraction of deskbar was that

- it was a commandline entry on the taskbar
- the menu with matches appeared right there, and was quite compact
- it was _all_ keyboard based, and damn fast

current deskbar has lost all of those items. You mention a refactor, and refactors are _fine_, but as a programmer who is right now in the middle of a major refactor, I'm sorry, but losing key functionality is _not_ ok.

I know you can do better - and with the new refactored codebase, it should be easy to add the commandline we all want, with the compact menu-of-matches and make that into a 2.20 release. Soon.

(If the refactor didn't make that easy... well, throw the refactor out the window, mate!)

Do it before it becomes part of the official gutsy. So far, only a small fraction of the deskbar fanbase have seen this.

In reply to by martin langhoff (not verified)

Do it before it becomes part of the official gutsy. So far, only a small fraction of the deskbar fanbase have seen this.
As said in my post, only the current UI will be available in 2.20. It's not because I don't want to support the entry mode. We are in UI freeze right now. Therefore, you're not allowed to change the UI without approval. I don't think that the UI is so bad that it's neccessary to request a freeze break.
Regarding keyboard navigation. I continued in improving it and you will see the changes in the next release.
I know most of you miss the entry mode, but I think the new UI is perfectly usable, too, and you got a great new feature (actions).

In reply to by sebp

At least could you change the fact it has window decorations, and attach it to the applet icon instead? (as it was before)
Nobody will for sure say "no!" to this change, and most people are asking for it

I am using the Deskbar in Feisty and I love it. I especially love the way that I can keyboard it out from the panel and search as I type (and you can almost do anything with it). Please do not change that. If the only way I can put Deskbar out is to use a mouse or to run it as an application, I will definitely use it less (I have a touchpad and I keyboard everything). I think your UI is really great in Feisty and wonderfully designed. If you have to change the UI and the entering method, please at least make it to be an option. Thank you very much.