Slashdot Mirror
Interview with Alfredo K. Kojima
BadlandZ writes "The UnOffical GNUstep site is featuring an Interview with Alfredo K. Kojima in it's new SlashDot-like [GNUstep newsWire]; news section. On a side note, I recently decided to publicly release a little essay I wrote on why GNUstep isn't just a DE for anyone who cares for a "non-technical" read on why GNUstep is a bit more expansive in design than say, KDE or Gnome. Just incase anyone cares... ;-) "
One point most folks don't remember about GNUstep: The end goal is to have one API with which we can write an application, and then just recompile it on any system and have it run. This means you write an application under Linux, test it out, copy it to windows and it should work without any tweaking. That's the end goal of this whole thing. Hopefully they will achieve that goal, because in doing so developers can support multiple operating systems with little or no work. Can you imaging what it would be like for porting if we could get adobe and corel onto this?
Don't forget, GNUstep's API will be compatible with that used by OpenStep and Mac OS X.
GTK+ has supported accelerators extremely well back to at least 1.0 (actually I remember using them in the 0.99s, but I don't remember when they went in). I would say better than anything else I've ever used. You highlight a menu item and hit a key and instantly that key is mapped to that item. If the app is programmed well (gimp is an example) it is even persistant when you quit and restart the application. No config files or menu editors to deal with, just highlight and click.
GNOME, last I checked, doesn't support this particular feature of GTK+ (at least the persistance part of it), but from what I've read in the UI standard that's going to be changed soon. I personnally think that this sort of thing should have been decided before they ever wrote a GNOME application, but I'm beginning to think that because of the way Bazar development works they may have made a better choice by going for applications first (in spite of conventional development wisdom). Maybe I'll write a paper on why in my copious free time, right after I finish that Palm Pilot program, make the recursive prefix modifications to PostFix, beat Zelda64, and re-read Lord of the Rings.
I seem to be the only one who can't stand the NeXT look, but I'm glad that there is that choice.. Now watch out for KDE v gnome v GNUStep.. :)
One of the reasons I switched my "allegience" from Gnome to KDE is that I found Qt to be a better match for my (nascent) programming style. I just wasn't comfortable with GTK+.
As I understand it, *Step uses Objective C. As may be ascertained from my flirtation with Qt, I like (ISO) C++. How much does ObjC differ from C++? Does it support templates? Are there good tutorials on the language and/or GNUstep programming?
These issues are more important than "this wm is better than that wm," "this library is more cross-platform than that library," or "this license is better that that license."
When will people learn: Mach is not UNIX. Repeat that. Mach is not UNIX. Mach is not even especially UNIX-like. Mach is a microkernel. UNIX-like "personalities" have been written which run on top of Mach. Personalities which run on top of Mach do not HAVE to be UNIX-like, though.
Why did we ever get KDE and GNOME anyway ? GNUStep was around (at least as a concept) long before KDE (and especially before GNOME). Why didn't the ppl. who wrote KDE didn't just use parts of GNUStep (and just write up the missing code themselfs).
.... It seems to be pretty moronic ..
Is this because the GNUStep maintainers are just too uncooperative or is there a diff. reson ?
Also, Why the hell did RedHat Start GNOME ? whyd didn't they just fiund Harmoney and get on with life
BTW, NeXTStep is alot more than just an interface you In theory have to modify a few things and u could get Enlightemnt running.
I don't know. Take a look at the GNUstep site; they also make mention of Apple's reference material, which includes an Obj-C overview. I don't know how good it is - I've perused it on and off for several weeks, but haven't tried even a HelloWorld.app as of yet. Maybe when those 28-hour days start, I'll have the time to...
--
--
=8^
- Page updated Nov 1998
...
...
- Site updated Feb 12 1999
- Try reading the GNUstepNewswire site for links to apps and projects
- Rob Current's article
http://rob.current.nu/next2000/
- the newsgroup/mailing list is quite active
- release (beta) 0.6 is due in March/April hopefully
Your "main beef" makes no sense whatsoever in this context: the *goal* is for GNUstep applications to run on all unices, Win32, and MacOS/X. And NeXT's UI (which it emulates) is consistent and quite useable - I mean if 90% of the world's computer users can get used to pressing on "Start" when they want to turn off their computers they'll be able to adapt to a NeXTish environment. Then again OpenStep apps compile with virtually a native UI in Windows so maybe they won't have to if GNUstep goes that route
I don't understand your complaint about open source. Currently 1000's of examples of open source software have been ported to virtually every existing architecture and OS (uhmm maybe you've heard of gcc, vim, Emacs, or perl?) and the whole point of KDE GNOME and GNUstep is to create more user friendly interfaces.
If it's a pressing concern then why not contribute your knowledge about HID to one of those projects?
Actually on closer reading your entire article is inconsistent and muddled and I've surely wasted my time so I'll sign
Anonymously yours,
Their progress page hasn't been updated in almost a year. I've heard hide nor hair of any GNUStep related project or application other than window maker and a host of related confuration tools and dock applets (which arguably, aren't gnustep related at all). It would be nice if development would continue, since OpenStep appears to be the superior API for GUI programming whether freeware or not.
:)). QT is a non issue, because I don't believe the windows version is open src, is it?
.feature.?? come on folks). GNOME can hardly be called cross platform for UNIX (just try to build this way over bloated system on anything but Linux, I dare you :)). KDE is better at this, but not remarkably so.
:)) and horrible human interface design (not something I can fault a programmer for, other than perhaps not making the leap to realizing that this design IS needed .. one only needs to look at the best example out there.. a wonderful program: the gimp, with what can only be described as a clumsy UI).
Also, if you look closely at GTK, you'll notice that while cross platform development is not tantamont in it's list of goals, it can probably do the job in some ways easier than GNUStep will be able to, using a platforms native compiler tools. (Hint: see GDK vs finding an objc compiler for windows
Both KDE and GNOME are pretty much laughable at this point. Obviously there is nothing better out there and that's a shame. The UI semantics of KDE blow (single click activation is a
Again, this is unfortunate, and these systems hopefully will, over time, get some clue infusion from people who know how to write cross platform code for UNIX based systems (it's NOT hard folks -- I would go so far to question the reliability of code written by people who can't do this), and how to design UI's usable by people other than the developers who wrote the UI.
I guess I'm rambling a bit, but this is my major beef with open source software at this point. Non portable (compiles, runs only on linux), untested (relying on the feedback cycle WAY too early in the process.. two words for these people: unit testing. See: glibc 2.1
I forget what its called but there has been a program around for a while which allows a user to switch between several virtual consoles even over a telnet session.
It would have been a lot better if Redhat invested in finishing GNUStep instead of making GNOME. Too late for that now... (just MO)
... that's your opinion and you are perfectly right to have one, but let me tell you a few things.
1) single click activation == feature
Damn right it's a feature!
What is the rational explanation for double click?
It only made sense in the mac because they only have one button. On anything else, it's silly.
I'm amazed by those who critizice KDE for being windows-like and then proceed to discard any difference without giving a rational explanation.
2) Objective C == proper language
Whatever floats your boat, but I believe the main reason for GNUStep development to be this slow -- and it's slow, people said GNUStep was going to be ready really soon in the original KDE thread, october 1996, and still there are no usable results -- is Objective C.
Why? Because only a few people know Objective C and most noone else cares about it.
It may be a damn good language, but if there is no developer mass, it can never make up for it.
I mean, it may be 10 times easier to learn than C++, and make development 5 times as simple, but if you have 1/100th the developer mass, you are still 20 times slower to develop.
GTKStep may be nice to some people (I don't like it myself), but we'll really have the chance for a well designed, completely open Destop Environment/GUI API in Warsaw and Moscow (parts of the Berlin project). I've taken a look at some of the design for Berlin and it looks really smooth; and it seems to have been properly designed *before* implementation (which may make it seem like development is going slow in the short run but in the long run it will benefit tremendously). Berlin's biggest obstacle will be getting people to think completely "outside the box" of X11. I wish them good luck.
--
Aaron Gaudio
"The fool finds ignorance all around him.
"Every man is a mob, a chain gang of idiots." - Jonathan Nolan, Memento Mori
Probably for the same reason we have vi, vim, elvis, emacs, xemacs, nedit, etc...
I think the first graphical Web browser ran only on Nextstep. this makes Nextstep look good because some pretty good computerists (the early implementors of the Web) decided to use it.
but where do you get the idea that if Nextstep had not been available the implementors of the Web would have just given up? do you have quotes from the early implementors?
Anyway, so much for devil's advocacy. ;-)
The "cue the foo posts in 3, 2, 1..." posts will commence with no subsequent foo posts in 3, 2, 1...
gotta say I just can't get into the way they work.
I dunno, I guess I just want a wm that'll get the
hell out of your way if you want it to and let you
easily juggle those dozens of sessions you have
open at any particular time.
So far, the SGI wm manages to do that best (after
some deal of tinkering, of course).
Don't mean to get too far off topic, but does
anyone else look at the new console in KDE 1.1 and
it's ability to switch between virtual consoles in
the same window and ask themselves, "why didn't I
think of that?"
----
Every year during my review, I just pray the words "slashdot.org" aren't mentioned.
Application developers want to reach the largest audience possible--not just people who are running the same windown manager.
End users want to just learn how to get around in one desktop environment.
I don't see any _nix taking over the consumer market without a standard "look and feel" across its major apps.
How the fuck can you call Windows ugly when you've seen something so horrible as NeXT? It's one of the worst looking GUI's I've ever seen! And how the fuck do they justify those 231231x123123 big buttons? are they afraid you might fit too much into one screen at a time?
There are no templates because they are not necessary with Objective C. Objective C has a type id which is kind of like a void pointer. So a container object that held ids could hold any kind of object. If it communicates with messages (which are fully polymorphic) then it doesn't need to know what it holds.
This means that all message sends happen dynamically at runtime, which is an overhead. However, these overheads often reflect a complexity in the program that C++ simply isn't able to do any better with. The result is a very flexible system.
Objective C shows its advantages best when you consider the ease of use and power of the libraries. It is also a strict superset of C, so you can use straight C for speed if you want.
That's just a random assortment of thoughts on Objective C, from someone who thinks C++ is a dirty word (so take it as you will).
...and applaud your efforts to disabuse the general public from their wildly inaccurate conceptions of GNUStep.
Readers take note: the rest of this post is not a troll (because I know someone will mistake it for one).
1. GNOME and KDE both suffer from the same disease: being implemented *before* being designed. Yes, both of them are touting CORBA-based component frameworks, but we all know that these frameworks sprung up *after* much application developement had already been done. Most GNOME and KDE applications are not component-based. They just look somewhat prettier and slightly more consistent than the average X application.
2. OPENSTEP has the real advantage of having been designed (in its original incarnation, as NeXTStep), implemented, used, revised, re-implemented, and used some more. So, it's not as if OPENSTEP is simply a design that looks nice on paper but can't be used to do real work. It *has* been used to do real work. The fact that it was never a hit in the marketplace is partially due to NeXT's longtime proprietariness and partially due to being tragically ahead of its time. NeXT was doing distributed objects before most people had ever heard of such a thing. Don't fall for the myth that OPENSTEP is outdated technology; the truth is, it's only now becoming timely.
3. Admittedly, GNUStep has some barriers to acceptance; it uses a language that few people know. And, more importantly, it simply doesn't have the momentum (or developer mindshare) of GNOME or KDE. But I think that part of the problem is that many people don't know what GNUStep *is*. And the article that "BadlandZ" wrote will hopefully have a positive effect.
4. BUT.. I don't think that he emphasized *enough* the fact that user interface is only one part of OPENSTEP. There are two major APIs: the Foundation Kit and the Application Kit. The former provides a basic, non-GUI, object-oriented framework for building applications. The latter provides the GUI classes.
5. I think that GNOME and KDE applications are consistent *enough* (considering the low standards of UI consistency in the marketplace). If you've ever used a NeXT box, you'd know that their apps are significantly more consistent. But, like I said, the others are passable in that respect, and that is not the prime reason to use GNUStep. The prime reason is simply that it makes application development (not just GUI development) an easier process. Granted, you have to buy into object-orientation, and you have to admit to yourselves that C is not a very good application programming language, however good it might be as a systems language. But reusable components are wonderful things when done correctly.
Hahaha! I love that sig11 icon in the screenshot...
i wish them all the best of luck in their project, however, the dream of having only one standard GUI interface is so anti-linux that i am amazed that someone will even bring it up. The reason that i love and use linux, is because i can CHOOSE what i want to use and how my system will look. There are so amny different choices, what they should work on, is a compatability standard, where everything can inter-communicate or something along that line. but what do i know, i only know what i like.
Its spelt "L-I-N-U-X", but pronunced as "Free Beer"
sorry, this is not true.
mach > 3.0 is a microkernel with several servers
available to run on top of it including a bsd
type. mach 3.0 was not a microkernel and was
a bsd monolithic kernel. for what it's worth,
the next did not use a mach kernel > 3.0.
At least Qt/KDE supports menu accelerators. Until you have to select something on the desktop without a mouse. Impossible.
Motif doesn't support them. (ever try to run Netscape out of the box without a mouse? Impossible )
GTK/Gnome does, in a way. Most GTK programs can't be used without a mouse.
Seems funny that all these developers are big proponents of a command-line, and once they go into GUI mode, keyboard interaction is only an afterthought.
Philippe
I'm inclined to agree...I prefer the look and feel of WindowMaker/GNUstep applications to the standard Motif and GTK apps. Qt (KDE) does a decent job of creating an interface, but compared to WM, KDE's just a hog...not to mention we've got the Dock and the Clip. :)
Agreed. I'm guessing that's why most people don't know about themes in openstep. Speaking of GTK, I'm curious if the design of the themes allows for NeXT-ish scrollbar buttons. The new version of GTKstep that uses themes has them back at either end of the bar. :-(
The "cue the foo posts in 3, 2, 1..." posts will commence with no subsequent foo posts in 3, 2, 1...
I suppose that you have to be able to look beyond your own individual GUI tastes and freedoms to see the intent -- which is not to impose One True(tm) GUI upon the geek, but rather to enable someone to easily impose One True(tm) GUI upon their users in a business. Like it or not, letting every drone in a corporate beehive apply personalized warpaint to their tools is a recipie for a disaster within the tech-support department.
You may be surprised to learn that this is actually consistent with a long tradition of consultants & IT departments that provide X11-based solutions to their users. The argument is that X allows you to set up a desktop where the only available operations are those that allow the user to use/maintain the applications that you provide, which is about as close to idiot-proof as one can get. In a setting where most of your users can barely figure out the semantics of mouse-control, this is a good thing.
NeXTstep found good middle-ground in the enterprise, giving end-users freedom to install local applications if they needed to, but not so much freedom that they could make life miserable for tech-support. The kicker: the GUI was designed so that it made an intelligent personal choice as well: the use of color only where it adds information, scroll adjustments that minimized mouse-travel, blah blah blah...
The "cue the foo posts in 3, 2, 1..." posts will commence with no subsequent foo posts in 3, 2, 1...