Slashdot Mirror
Building Nautilus: Behind The Scenes
mholve points to this LinuxPlanet feature titled "A Sneak Peek at Nautilus from Eazel." Despite that title, it's not just a collection of bulleted feature lists and screen shots. Instead, it's a nicely balanced, in-depth look at the thought process behind the creation of Nautilus as well as a description and critique of the current preview. The article raises some interesting points about the complications that Eazel has found in trying to make Nautilus distribution-neutral: "In the future, I think we're going to have to look for a way to abstract the package system much in the same way we've abstracted the file system," [Darin Adler, Eazel software engineering honcho] allows. In addition, he notes that the diversity found in other areas of Linux distributions presents a daily challenge: "We try to get weirdnesses out of our code as much as possible, when we discover we've accidentally done something distribution-specific, we try to 'unweird it.'"
Ironically, their action probably had the reverse effect. OO code in C++ is better maintainable than OO code in C. The ++ in the language name means that some usefull features were added to the C language. No (or hardly) any features were taken away.
It's a pitty that eazel went to work with gnome instead of kde. At least their developers recognized kde it as technically superior. This is extra sad now that the license issues are resolved. Luckily the KDE developers didn't wait for somebody else to fix their problems and created Konquerer and KOffice.
Jilles
I don't understand why the free software camp is prepetuating that legacy. If Microsoft had had reason to write a file manager/spreadsheet program, say, or a file manager/photo editor, does that mean that the free software world should continue the same strange groupings?
It doesn't strike me as good software design to write a single program which performs two unlike functions. Better to write two separate programs, and then, if there's a need, work out a way for them to talk to each other, while keeping the internals of the two completely separate.
Um, ah. Parse error. You said "GUI like GMC", or words to that effect. That clears it up, at least somewhat... What you want is a dual-paned file manager but with icon-based display, right? Cool. Good luck finding one.
- In a per-directory file, e.g.
.iconinfo or something. Perhaps the best choice on typical Unix-y filesystems...
- In a global magic database somewhere. Keeps individial directories tidy, but is probably difficult to get to work well with (dynamic) mounting etc.
- In the icons themselves. This was how the Amiga did it; the Workbench only displayed icons for files that had a matching
.info file (so foo's icon would be in foo.info). Works well, but creates an extra file for each visible file. Uglifies the filesystem, and takes loads of space.
- Not stored at all. This seems to be how Windows does it, I think. Icon positions are only maintained as long as the particular window is open. Clearly, this doesn't solve the problem (for me).
;^)/
Wow, I guess I got a bit carried away, there. Anyone got any ideas about this? Let's hear them!main(O){10<putchar(4^--O?77-(15&5128 >>4*O):10)&&main(2+O);}
A file manager is for managing files
One problem with the managing of files is that, in general, users don't want to manage files and directories, they want to manage information. It's only because we've had years of experience with computer systems that we can effortlessly make the file -> information jump.
So, the real problem is how to build an extendable interface which allow people to build "information management" applications into their user interface. In this light, a web shell is actually a decent design decision -- it's runtime interpreted, the language is widely understood, and you presumably have the software installed anyway. (The downside is local vs remote security.)
Aside from the political and legal problems with MS's IE integration is the fact that they never did a really good job providing any sort of decent "information management" on top of explorer. (JPEG preview is merely the obvious gimme.) So, like you say, in a stock Windows config the reasoning behind using a web component is really non-obvious when you can have a simple OSX-style browser
Now, Nautilus, from what I can tell from the screen shots, really takes this up a notch with far more component integration. Whether or not this is a useful way of managing file-based data remains to be seen. But, hopefully it will be a full enough implementation that (unlike Windows), users will get the idea and start runnig with it.
What I'd really like to see, for example, is a hierarchical search interface which abstracts the actual directory structure. Also, a standard policy way of doing database-like attributes is a longterm must. "Home pages" for shared project directories and integrated versioning would also be features that I could start using tomorrow. Something like Natalius would be the perfect place to start to implement these sorts of interfaces.
(Note that I'm not trying to say that something like Nautilus is a cure-all, just something useful. There will always be the need for a shell, a stripped down Mac/Win95 file browser, and specialized database and cataloging tools.)
When I hear the word 'innovation', I reach for my pistol.
Even if he uses emacs ... you can probably get to vi through a shell. But on a Mac or Windows box, you're not going to find WordPerfect if the owner is a Word user. Even worse is if you need to do some image processing (for example) and he or she isn't a graphics person; then you're not going to be able to do certain tasks at all.
This is actually one of my complaints about Linux installs. You try to do a basic RH install and you'll be getting 650-750 meg of stuff, most of which you won't ever use.
Why should I have to install 12 editors if I only use emacs (with the occasional jump into pico)?
And it's not just editors, it's across the whole spectrum. I've tried to do installs where I select only certain components, but then I always have dependency problems that make absolutely no sense and I STILL end up with garbage that I know I didn't select being installed for me.
Why can't we easily get an install that works correctly with nothing more than a kernel, the gnu tools, x, a desktop and window manager of choice, and a couple of basic apps?
If Microsoft came up with a distribution of Windows that came with all the crap that Linux distributions come with, we'd all complain about it.
Maybe this was ok when there were only a few apps available for Linux and people just chunked everything in because they could, and so they could prove that apps for Linux actually existed, but not now.
I just don't understand why when I want a non-Xed server that due to dependencies for text based apps I am required to install XF86 libraries?
load "linux",8,1
This is a problem of education. Certainly there are many users who simply want to get work done, but understanding the mechanism versus policy issue would help many proto-geeks to gain a better understanding of why X is the way it is.
GNOME and KDE are certainly good places to have policy.
The X Window System is not; if you warp X so that it enforces policy, you will have an X that many current users will not or cannot use.
--
"Where, where is the town? Now, it's nothing but flowers!"
Cheers
Everything is but a number spoken by itself.
I've never used it so i really can't say. But I do know that the main dude for it worked at Apple for a while. so I imagine he wasn't really trying to copy Microsoft UI. He talks about how they carfuly chose when to borrow, and when to invent. hopefully he wasn't just blowing hot air.
A file manager is for managing files. A Web browser is for viewing files. To integrate the two is to create confusion
Not unless your a complete freakin newbie. I think most people understand then when you type in c:\temp\ (or /root/) you'll be be in local file mode and when you type www.slashdot.org you'll be at a website. As for this being "bloatwhere" it's all in the implementation. I'm sure your aware that explorer ISN'T IE, it's just a wrapper for IE, even IE is a wrapper for IWebbrowser. they are loaded at different times, it's not like they are both loaded in RAM when you load it.
-Jon
this is my sig.
One thing I've noticed is that most of the file managers nowadays go on and on about how good their graphical interface is, how well implemented drag-and-drop is and so on, but almsot none of them touch on how the keyboard can be utilised.
I understand that some people would prefer to use the mouse to do everything, in which case something like Nautilus would be ideal, but how about those of us who prefer the keyboard? Adding in a good set of keyboard mappings is quite easy to do, but sometimes its sorely neglected.
I almost never use gmc. When I want to work with the filesystem, I open up an xterm. When I work in Windows, however, I use Explorer all the time, and one of the reasons for this is that its very easy to do everything with the keyboard. I can select a set of files, Ctrl-x them, Alt-Tab to another window, and then Ctrl-v them into the new directory. To me this seems much faster than arranging the windows and then dragging the set of files to the other directory.
One reason why I don't like dragging and dropping with the mouse is that I find it a bit stressful on my hand after doing that for too long. If somebody has a physical disability that prevents them from using the mouse in such a way, can he/she still use Nautilus?
Does anyone else think this looks *exactly* like Windows Explorer? Particularly with the revised design in Windows 2000? Where's the innovation?
- I don't care if they globalize against free speech. All my best free thoughts are done in my head.
I think it's good that Easel chose to work with the Gnome framework, but the reasons given in the interview left out a lot. The real reasons are more like:
1. Kde already had a similar product, Konqueror, well under development at the time. Konqueror is now in release state or very near release and is quite stable, fast and fairly efficient with memory use. While Kde always welcomes more developers, I don't think they would have wanted developers working at cross purposes with Konqueror so near to completion.
2. Kde and Easel have different philosophies. Easel is a money-making project started by entrepeneurs hoping to capitalize from application delivery and commercial services using the Easel/Gnome framework. Both are free software, but Kde's purpose is simply to provide needed free software to unix users, not to make money from such software. Gnome, through Helix and other spinoffs, already had established itself as a free desktop environment for the purpose of such capitalization by developers and entrepeneurs close to the project. This more than anything mandated Gnome for Easel.
3. Gnome did not have a competetive file manager/browser in the eyes of most users. Personally I like gmc *a lot* but it does have its limitations. Therefore Gnome had a real need for Easel.
4. Of course the licensing issue with Kde was a factor, and the only reason given in the interview which was legitimate. Not that the Eazel principals thought that there was a problem with licensing (they didn't) but the uncertainty among others made Kde a less attractive choice at the time.
I'm a big Kde enthusiast - using Konqueror right now. However, the competition from Easel will help Kde. Of course I think Konqueror already provides most of what Easel promises, and much more, although in some areas Easel will be innovating first. Integrated web/network/local search comes to mind.
The main advantage of Kde 2 is efficiency and speed, both for the user and developer. For example, Kde 2 is performing beautifully in most ways with my now antiquated P200mmx 32 meg PC. When it doesn't I bitch at the Kde developers and lo and behold they fix it the next week. "Your Kde is in deep shit! Now fix it!"
Everything, or almost everything, in Kde 2 really is a component either with a thin wrapper to make it a separate app or embedded in another app. It works smoothly. I think Gnome will have a lot of trouble getting that working without tremendous bloat and slowdown and stability problems - when it is *really* used by the vast majority of apps like Kde is using components right now, not just by a few showpieces or to embed a memory monitor into the panel.
I hope Kde learns something from Easel about the importance of a "finder" to users. It should be fairly easy to integrate that into the existing Konqueror with compnents and the results can already be displayed and manipulated in a number of ways with Kde 2. (Konqueror already has multiple paned browser windows and the ability to embed different views or viewers in different panes if you want to, though I usually prefer one or two panes).
Finally, on the choice of languages, if CORBA and bonobo really are language neutral, then the preference of the Easel developers for C++ should have been respected. But these interfaces really are not so language neutral after all. So, I don't think it was so much a matter of Easel being "pressured" by Gnome founders to use C with Gtk+. It was a matter of their realization that C with Gtk+ would be more likely to work with the Gnome framework and that C++ would cause problems for them.
Sure, you can use languages like Python and Perl to glue components together, but the components themselves usually are written in C, or C++, depending on whether Gnome or Kde is being used. It is possible to use other languages for creating components but no so easy to use anything other than C with bonobo or C++ with Kparts. CORBA is certainly not the magic bullet though it has uses in some situations.
I do know from personal experience that application development in C++ with Qt is a lot more FUN and faster than using C with Gtk+. This is especially true for more complex apps. I can imagine the cursing and swearing at the Easel workshop where coders are trying to use a mixture of Gtk's gimped up object oriented C, corba IDL, and a constantly changing set of libraries requrired to support Gnome and Gtk.
Anyway, I am quite impressed with what I have seen of Easel. They have done much in a short time under trying circumstances as explained above. Congratulations to the Easel coders and designers!
I understand the sentiment, but there's a difference between "spartan" and "highly usable." I think Gnome, et al, are going off in the wrong direction. They're interfaces for the sake of being interfaces. Arguably, that's okay, because for many people Linux is the operating system that exists solely to be an operating system :)
So, yes, I'd rather see less cruft devoted to customizing interfaces and fiddling around with a desktop, and more focus on real tasks. But if you take away flashy GUIs from Linux, you don't necessarily end up with more. You end up with something slimmer, faster, and more reliable, but often awkward and misdirected. The next step is to design something that's geared toward what people need out of computers. Jef Raskin has done much work in this area. The OS in the Apple Newton is another good example; it's more radical than most people realize. Open Genera is another OS designed with real problem solving in mind. Unfortunately, we're not seeing much in this direction any more, as being like Windows is seen as much more important (kinda surprising from the subversive Linux crowd, I must say).
Navigation bars should always be on the "left" its a pretty safe bet to assume most people will feel much more comfortable dropping menu's down from left to right (english speaking folks) anyways.
Think about it, you read left to right most everything you do is from left to right, typing its deeply ingrained in your mind by the time your 20 or so.
Going from right to left seems to make less sense to me, so do sights that put the menu on the left just to be cool?
It dont make any sense to me.. Thats why I like left to right navigation on all my sights as I think it makes it more useable to a greater number of people...
Jeremy
What's this about the Gnome project strong-arming the developers into switching from C++ to C? Leaving aside my religious beliefs in OOP, that doesn't seem in keeping with "open source ideals" at all.
Worse, I thought gnome was supposed to specifically address inter-language issues (everything thru an ORB, or whatever). The fact that this issue even arose suggests that there is some kind of deep architectural unsoundness to this project.
This wasn't going to be just another module, this was going to be a part of the core GNOME...
A file manager is part of the core? Must be written in the same language? Bad signs. Someone who knows better tell me I'm wrong.
Kill, Tux, kill!
I'm just going to say what we are all thinking. Well half of us anyway...
Some people like stuff like Nautilus, evolution, Gnome panels, really loaded root window menus, etc... Others like about 4 items in thier root window menu, gno panels (and I mean gnone), run emacs (as file manager, editor, ftp client, eye-washing area, short-order cook, etc), and just think that click-heavy interfaces slow them down.
As someone who prefers lightweight interfaces (text) in a slimmed down windowing environment without gnome, etc. I can also appreciate the attraction of a richer interface experience. But since I am not into gnome, etc, I am not going to say anything for or against it other than to thank all the developers and testers out there for all their hard work.
I do have a question though. Most of us who have been using a system that is highly configurable and application rich (I don't want to say "Linux" because there are lots of others) settle into a rut after a while with respect to the tools we like and how we want to have them layed out---whether we like lazy focus or click-to-focus, auto-raise, emacs, vi, etc.
When I sit down at my friend's Linux machine it's a bit wierd. Nothing works exactly like it should and I don't know where to find anything :) then I sit down at a Solaris box and the same disorientation follows. I prefer this to being in a monotonous environment but I want to ask if Windows users experience this same mild disorientation when they use other boxes than the ones they are normally on (do other unix users or am I am freak)? I am pretty sure Mac users do (again due to it's high configurability).
--8<--
--8<--
I am glad to see the progess in Gnome. Will the new gnome create a major shift in the Desktop war? I hope so I hope to see that KDE and Gnome gives choice to the consumer. It choices that helps make things better.
As a windows user, I can see what you're tlaking about, and I have the same issues with other people's windows machines, though not to the same extent. For the most part, you can trust the Start menu to have the same 8 (if you're using win95) initial items. Anything found above that can probably also be found under Programs.
;D ). Radically different cursors can give me that same, pseudo-subconscious feeling that something isn't right. And when you have things with panels that slide out when you nudge the edge of the screen with the mouse, the same thing happens.
But there are little things. Color scheme, though that's not much of an issue. Font size (under Control Panel->Desktop->Settings) can give me a sort of vague feeling of discomfort, because it messes with the maximize/minimize/close buttons, till I eventually break and set it all to Small fonts with 1024x768 (then I restore it when I'm done
But it's the same with every OS, I think. I'm used to my customized DOS prompt ( LCARS | C:\> in bright green, so the basic C:\> in light gray is kinda boring). I'm used to the way I have apps arranged on my Palm. From what I can tell of Linux, though, you have a much greater degree, or at least ease, of rearrangement and customization.
-J
Karma: T-rexcellent.
The article is excellent. The topic most important to me on Slashdot is the progress of Open Source development.
It's interesting to get a glimpse into the inner workings of what apparently is one of the most socially important and socially advanced endeavors in the world today. Open Source development is an example of humankind working in a truly cooperative way for the benefit of all of us.
Bush's education improvements were
None of those things have anything to do with X Window System. To say that they do implies a misunderstanding of what X is.
Simply put, X provides the mechanism, not the policy. X is plumbing and electrics. X is not your interior decorator.
I don't know what version of GNOME you last used, but GNOME 1.2 runs like greased shit on my Pentium II box.
Quake has nothing to do with GNOME. Frankly, if you're running a game on anything but a bare X server, you're already on a loser.
I'm using nightly Mozilla builds as my main browser and have been for three months. Your estimate of a year is unduly pessimistic.
If you find that Windows 2000 does everything you want out of a desktop OS, then by all means keep using it. You won'e be struck down by a bolt of lightning if you do.
--
"Where, where is the town? Now, it's nothing but flowers!"
It's good to see some really experianced UI developers doing something with Gnome. I really wonder if they will fix all the little things that don't work well in X, things like the wheel mouse, TAB stops, consistant hot heys, even window focus - much bigger issues too, like an consistant way to install programs. an easy way to configure all my hardware (which drivers for will always seem to be in beta). and most of all SPEED Gnome makes me think my computer is a year older then it is. hell i can't even play Quake III on it without crying.
I'm a little out of the scene, are these already working well? Last time i used X I was an enlightenment or blackbox junky. I always found KDE and Gnome to be cheap 1.0 like rip offs of Windows, the really big important things worked. but all the little details we're missing.
Truth be told, I'm not going to use and *nix for a desktop until i get a decent browser. Mozzila seems to be coming around nicely. maybe in a year of so. So for now i'm sticking to Win2k.
-Jon
this is my sig.
This is no different than the closed source model, except that the end user DOES have the option of fixing the bug him/herself, if they choose. But in my experience they tend to only do this when they absolutely need to (original author(s) don't respond, or are dead, or something).
Of course, there are nice people out there that will once in a while submit not only a bug report but the fix as well...But this is a hard thing to do, because finding the source of most bugs requires a pretty deep understanding of how the code is working, which can be very time consuming.
Well, the short version of all this is that while many eyes DO make all bugs shallow, those eyes have to have brains attached that are capable of understanding the code and also enough free time available (and the desire) to learn the internals of the buggy program.
They said the Gnome core developers wanted to be certain they would be able to take over if Eazel crapped out. Makes sense they would want it written in the language they prefer so they could maintain it.
Well, yes. They are trying to make a warm-fuzzy-gui "desktop environment" to make all the hordes of windows and mac users feel at home, after all. This stuff is NOT being written for the hackers. Given their goals, it makes perfect sense that the file manager be considered part of the core. Gmc is certainly a central component of gnome now.
=-=-=-=-=-=-=-=-=-=-=-=-=-=-
Friends don't let friends enable ecmascript.
I've stopped using GNOME a long time ago to replace it with just plain enlightenment. I've recently started using EFM, the enlightenment file manager. While EFM is still pretty slow and buggy (version 0.0.0), it has many features which make it unlike KDE and GNOME's filemanagers. For example, if you begin typing in a window, a transparent box will come up so you can see what you're typing, with this you can do all sorts of things, from selecitng files to running commands to opening web pages. If you were to type 'www.slashdot.org' it will open a new NS window of it, typing 'make' will launch an Eterm running make in the current directory. EFM fits my needs better than any of the other linux FMs that I've seen, annd I can't wait for it to get more stable. Though nautilus has many interesting features which EFM doesn't, they're probably aimed at different audiences anyway, EFM is more 'graphical shell' than filemanager. Nautilus may be good for people who want an embedded web browser/mp3 player/etc. in their filemanager, but I think that we can all benefit from their work, regardless of weather we are interested in the filemanager itself.
Shine on, you crazy diamond.
You weren't just judging based on the looks of screen shots, did you?
GNU/Linux. The Freshmaker.
In every case: navigation buttons across the top; in the main window on the right, the object you're viewing is rendered - jpg, directory, info file, web page, whatever; on the left a panel gives you control over the object. Great: stuff everything into the same model as far as it will go. I don't see any blatent misdesign so far.
Is there a file edit view? Can I gdb an elf file, then <back> into the editor and <forward> in the debugger?
--
Life's a bitch but somebody's gotta do it.
Nautilus is certainly a fine tool, and a very good copy of Microsoft's IE-integrated file manager. But even the best copies of a bad interface are still bad interfaces. There is no need for quite a few of Nautilus' features. Foremost among these if the integrated Web browser (Konqueror is guilty of this too). A file manager is for managing files. A Web browser is for viewing files. To integrate the two is to create confusion, because you have the same app doing entirely different things (remember the OSX Dock, both an app launcher and task switcher? Same basic problem, though the purposes it serves are different).
Gnome doesn't need The App That Does Everything (neither does KDE, for that matter). All that does is lead to bloat, redundant apps on a system, and confusion for new users.
The Win/IE filemanager does have a few strengths, but these can be gained without bloating the filemanager itself. For example, the ability to show file information in a sidebar, rather than pulling up a new window for it. OSX's Finder can do this without the need to throw a browser in. The ability to navigate a filesystem using buttons in the toolbar is a Good Thing (particularly if you use only one window for file browsing), but you don't need a whole browser to do that (OSX's Finder, however, does this one very poorly; just a Back button which runs totally counter to the column-view paradigm).
I'm not trying to bash Gnome or KDE. I'm just pointing out that they're making quite a few of the same interface mistakes Microsoft made, and it'll only hurt them in the end.
----------
The Stop button (big X) should be just to the right of the Reload button, not all the way over on the right side. I can see the thinking is 'it's like a period, put it at the end of the sentence'. But I think this is superceded by: I always want it in the same place, where I can find it fast, near the other navigation buttons. Having it on the right off the location pre-sets messes this up.
A more important issue: suppose the Nautilus designers and I disagree on exactly where the best location for the Stop button is. How do I put it where I want it in *my* version?
--
Life's a bitch but somebody's gotta do it.
Ehum, would you mind elaborating a bit, there? I don't really see what adding a GUI front end to gentoo would mean in practice. In my world, it has a very graphical front end, albeit one dominated perhaps by rendering text. Hm. Are you after a more "icon-dominated" view here, or what? Curious developers want to know! ;^)
main(O){10<putchar(4^--O?77-(15&5128 >>4*O):10)&&main(2+O);}
1 2 3 4 5 6 7
The home page for Eazel here.
...and their latest screenshots are here...
Why is it that every graphical file manager only has one file pane? Why is it, that GnomeMC is nothing like midnight commander? It doesn't have any of the features, or quality that midnight commander had. It doesn't have one tenth of the features, or usability. I shouldn't even be allowed to take the same name!
;)
Its seems that we have another dull file manager here, one meant for people that don't want to manage files, but just look at them, and access them. A single pane interface is slow for copying, and working with the files. Sure, it looks nice, but I'll take something like Filerunner or Gentoo ANY day over something that looks good, but is completely non-functional.
When someone takes something like Filerunner or Gentoo, and then gives it a GUI front end like GMC, then I'll be interested. Otherwise, what's the point?! I can use Filerunner right now to do everything GMC does, plus tons more!
Bah! (yes, this was a rant