Sunday, February 28, 2010

Nautilus in the stress field between design and function

So, there's been a lot of talk about Nautilus' present and future design. Especially Izo's article collects some interesting proposals for design enhancement that are in my opinion very well worth consideration. It does, however, also contain some points that make me feel a little uneasy.

Naturally, UI redesign comes not only with re-organisation and re-design of existing functionality, but also with omitting of unimportant components. Also naturally, the latter part is tricky to get right, because people don't like it when the functionality that they are habituated to use is suddenly gone, without any accessible replacement. I don't mean to say that UI reduction is impossible, but it's something that has to be considered carefully, according to the focus and target audience of the project. This is a hard process, because individuals or small groups don't necessarily cover all the use-cases that an application wants to address.

Izo seems to have a typical case of forgotten addressed use-cases in his review when he writes about the stop button:
"The stop button. More useful in web browsers, if you want to stop the web browser from loading a page, completely useless in a file manager, where file accessing times are considerably quicker than web browsing times. You simply never have an opportunity to stop the file manager from loading a page. It’s an old relic. I’ve never used the stop button."
That you have never used a stop button in a (network transparent!) file manager does not mean that it's "completely useless". Got the hint in the parentheses? Nautilus does indeed want to support network filesystems, be it NFS and friends or GIO/GVFS mounted stuff. Loading these is actually not too different to what you do with your browser. These can clearly be very slow, and that you didn't think of that just shows that you forgot an important use-case that Nautilus wants to support, because you have no personal use for it.

Now, I am not saying that the stop button is the best design for that. It most probably isn't. One could for example test if it could be combined with your proposal of the "refresh" button: Display the "refresh" symbol when the displayed location is fully loaded, and put a little stop symbol there while it's loading (this combination of stop/refresh has been proposed before in "Simplified Nautilus").

People are different, and have different ways to use applications. Many other proposals, for example, just assume that the sidebar is visible. You mean to save a little bit of screen space by omitting the small "Home" toolbar button, arguing that it's also accessible in the sidebar? Well, when I look around, a fair number of computer-novices that I see don't have the sidebar visible at all. When forcing it upon them, they actually loose a lot of screen estate, and have a lot more unwanted UI elements in their face. Combine that with the talk about removing the menu bar alltogether (and thus loosing the habituated easily accessible way to toggle the sidebar visibility), and you have a good potential to regress usability for a non-negligible part of your target audience. In the end, people don't open Nautilus to lay back and enjoy its look, but to get work done.

I am also not entirely convinced that it's so bad to have two toolbars. I mean, how much screen estate do you loose, in reality? For one thing, the toolbar covers less space than the sidepane. In browser mode, you typically don't have dozens of Nautilus windows on a single workspace (and as I said, the toolbar covers less space than the sidepane). On the other hand, even if you have reduced the number of toolbar buttons to six plus a wide-enough search window, as your mockups show, what's better: To use a few vertical pixels, or not to be able to see your current folder "Pictures from Patrick's wedding where aunt Maggie got really drunk"? Hard to say.

Split View - Curse or Blessing?

Having the location bar embedded into the toolbar items also has another problem: It doesn't work well with the a split-view filebrowsing mode. This mode got some very harsh criticism at the recent designers hackfest that I stumbled upon by coincidence (strangely, this discussion didn't find its way to Nautilus communication channels yet). I wrote some comments on the corresponding blog posts, but I also want to write a small comment about that here on my own webspace, without the risk of being moderated (as seemed to have happened on other people's digital homes, which obviously found my remarks unpleasant).

First of all, I am disappointed by the way the criticism was expressed. The designers may very well have some valid points, but I wouldn't know, because they seem to actively refuse to answer my request to elaborate on their non-descriptive slating. Also, there are some remarks that make it look like they didn't even give it a fair try before condemning it.

A reccurring question was "why split-view", and the proposal to implement panes in the window manager instead. This is a valid question, and has in fact been discussed on the Nautilus mailing list. (Hint: In general, project mailing lists are a good place for both, to research design decisions and to ask the "why" question.)

Fact is, split-view filebrowsing is not trying to solve window manager shortcommings on the wrong level. Even if Metacity had snap today, this wouldn't be an alternative.

The key difference is the inherent connection between those two panes, which gives clear benefits. When you want to do file management (which is now becomming the key-focus of Nautilus), you often have to deal with two locations at the same time: The source, and the target. What split-view does is to display a "default target" right next to the source. That's why it makes sense to have 2 panels, but not 3 or 4.

This default target is accessible in the menu, via the "{Copy| Move} to other pane" items. Users that need to do heavy-duty file handling can assign keyboard shortcuts to those menu items, and move around files with a single button press.

The “default target” notation offers even more for advanced users, which can access both the source and the target pane in their Nautilus Scripts (and write for example a “diff these two directories” script with very little effort). This surely isn’t possible with a WM snap either.

Users with simpler needs aren’t really affected much by of all this. The only effect for them is that “extra pane” option in the view menu that they don’t wanna click.

I actually think that it is a very natural model to show source and target location when they are that fundamental to the typical action that the user wants to do with a given application. I find it much more intuitive than the clipboard copy/paste stuff that is generally accepted because people got used to this strange idea.

Photo © Adventures in Librarianship on flickr, cc-by-nc-sa

Friday, February 26, 2010

Linking notes and email messages

A few days ago, I've cobbled together a note-taking solution for my email messages. It's very unixy, consisting of a fair number of different parts working together: Claws Mail with the Python plugin and two of the shipped example scripts on the one side, Tomboy with the Claws Mail addin and the Reminder addin on the other side.

Starting with a message selection in Claws Mail,

a click on the "Create Tomboy Note" menu item of the Python example script results in this dialog popping up (I know that this dialog is ultra-ugly, but hey, it's a quick&dirty easy-code example script, nothing more)

which in turn creates this Tomboy note

The Tomboy reminder plugin will take care to remind me about this next monday by raising the note.

To make the round-trip complete, there's a second example script that raises all Tomboy notes that link to a selected message.

Okay, admittedly not the end-user friendliest setup, but it suits my needs pretty well. And it shows the benefits of scripting language interfaces for glueing components together -- it didn't take long to write the scripts to make this work, even though the plugins and addins have not especially been designed for it.

Monday, January 18, 2010

The Tomboy and the Git

Tomboy's taming of the beast turned out to be a very useful feature for my daily note keeping. But emails are not the only pieces of information that I often find associated with tasks. Another recurring source that I want to reference are commits in source code management systems. So, if Tomboy can play nicely with my MUA, why shouldn't it play nicely with my source code repository browser as well?

Unfortunately, there's no drag-and-drop target for git repository viewers defined. Most viewers just don't let you drag from the commit list into another application. So, I tried to contact the guys from gitk, giggle, and gitg in the hope to define such a dnd target. The guys from gitg seemed to be the only ones interested in that functionality (good thing that gitg is currently my favorite browser anyways), and it didn't take long until they added the required features.

With that in place, it was easy to write a Tomboy addin that handles dropping of git references into a note analogous to dropping email messages: By creating a link with a nice icon and a meaningful text which when clicked opens the git repository viewer and selects the respective commit.

I like these little helpers. They have a good work/gain ratio.

Sunday, January 17, 2010

Extending and Automating Claws Mail - the sneaky way

The recent release of Claws Mail 3.7.4 has also seen a much more powerful version of the Python plugin. It is now possible to write scripts that are executed automatically on startup, shutdown, or opening of a compose window. It's also now possible to write scripts that work on an already opened compose window. The user interface got better as well (e.g. it's now possible to trigger scripts via toolbar buttons).

However, what the latest release still lacks, is documentation and examples. After all, features that are not documented don't exist. This is supposed to get better in the next release. I've started adding a few example scripts to the source distribution that show possible solutions to questions that have been raised on the user's mailing list lately. Most of these should already work with the released version of the plugin, with the exception of the startup script that show's how to add new menu items for custom actions into the main window (the examples being a menu item to mark a thread as read, and to add a menu item to create and show the Python plugin's API documentation on-the-fly - isn't introspection cool?).

Anyways, if anybody scripted something cool with the plugin, please consider sending the (well commented) script to me. I'd be happy to consider it for inclusion in the distributed examples.