[CS-FSLUG] Linux Today - Is desktop Linux too fragmented to succeed? A friend tells it like it is!!

Timothy Butler tbutler at ofb.biz
Tue May 5 11:13:11 CDT 2009

>> 	How do KDE 3 and KDE 4 apps interoperate? For example, I can run an
>> app written for Mac OS X 10.1 and one for 10.5 and not know which is
>> which.
> I can also run KDE 3.1 and 3.5 apps and not know which is which ;)  
> (just
> a joke...)

	Yes, yes. I'll admit OS X's version scheme isn't the clearest. ;-)

>> Likewise, I can run many Windows XP, or even 3.1, apps on
>> Vista and not have a clue which is which.
> I did not use Vista that much, but for the looks I can tell whether an
> app is a Win 3.1 app, or a Visual Basic app, or a .NET app, or a  
> Borland
> Delphi app (the icons on the buttons are a dead giveaway for the
> latter).

	Right, some of them will give themselves away with dated icons and  
such. But none of them is a "second class citizen," with a slight  
exception in the 3.1 apps. Dialogs and such will work exactly the same  
in an early Win32 app as the latest 64-bit, Windows 7 app.

> Then, after running KDE4.0, I had two completely different directory
> settings and I could even switch between the two environments. And  
> while
> using KDE4, I could use any KDE3 app. But I recognize this is major
> surgery...
> My point is that it is doable.

	This is where I'm wondering if KDE couldn't help out. Really, it's a  
bit of a waste to install the whole DE just for compatibility's sake,  
but more importantly, now you have two sets of configuration files, etc.

>> 	I think the big thing, though, is that KDE does keep breaking the
>> ABI even if you can kludge around it.
> So what? Gnome did it. Mac OS9 -> OS X? Apple did it, too...

	Right. It's fine to break the ABI. But KDE has four separate ABIs:  
1998, 2000, 2003 and 2008. Apple has two over a much longer period:  
1984 and 2001. GNOME also has two 1998 (ish -- GNOME betas were in  
fairly wide circulation, perhaps 1999 is fairer) and 2002. GNOME 3  
looks like it will not break compatibility.

	Note that Mac OS X, Windows and GNOME all seem to follow a similar  
philosophy: legacy support is slowly phased out as needed, which  
causes some older apps to fail, but they don't (as a rule) break the  
entire ABI. It's a gradual progression. I think that fits peoples'  
natural progression with apps too -- eventually, people upgrade, but  
one might upgrade Office in a slower fashion than Quicken.

	Mac OS's big ABI break was a virtual necessity; Classic simply was  
too far behind and attempts to upgrade it directly had failed numerous  
times. But, even then, they went to great pains to insure that the  
Classic emulator integrated the old apps as well as possible for half  
a decade after the OS X had come out.

>> Mac OS X has not broken its
>> ABI. And, when they switched architectures, Apple went to great
>> lengths to hide even something as complex as an emulator or another
>> architecture (!) so that the user never sees it.
> I think this is because they really made the mistake of switching
> architectures... IMHO they had the advantage with POWER.

	I think they did too, but the problem was the inability to secure a  
good selection of PPC processors oriented towards the desktop. Since  
the Intel transition, Apple has been able to move at a much faster  
clip; while RISC has some inherent advantages, Core Architecture I  
think would have trounced PowerPC in benchmarks even if IBM had  
slightly improved its speed of developing new chips.

	In the end, Intel Macs run just like PowerPC Macs do, only much  
faster, so it hasn't been much of a loss. And not having to explain  
differences in processor architecture to people in marketing materials  
is probably nearly worth it by itself for Apple.


Timothy R. Butler | "Bad is so bad, that we cannot but think good
Editor, OfB.biz   | an accident;  good is so  good, that  we feel
tbutler at ofb.biz   | certain that evil could be explained."
timothybutler.us  |                           -- G. K. Chesterton

More information about the Christiansource mailing list