Fog Creek Software
Discussion Board




Kylix development abandoned

Borland announced that there won't be any updates for Kylix in 2004.

http://www.linuxworld.com.au/index.php?id=122384005&eid=-50

This tells us a lot about Linux. It tells us that it is unlikely for commercial applications to succeed on Linux.

So, on Linux there will be 2 kinds of work:

- write code for free

- tech support for existing code

Please, kill me now!

John
Friday, November 07, 2003

There have been no spaceships flying to the moon in the last three decades. I guess sooner or later the lights will go out on the moon.

Johnny Bravo
Friday, November 07, 2003

"It tells us that it is unlikely for commercial applications to succeed on Linux"

Yet another case where Logic should be taught every year all the way through high school.

Agree with your conclusion but not how you got there
Friday, November 07, 2003

Take the experience of one poorly done but expensive development tool and extrapolate that to all software. Right...

(note to Delphi fans, I'm talking about Kylix being poorly done - Delphi on Windows is an excellent product, Borland just did a bad job of the linux version)

RocketJeff
Friday, November 07, 2003

It may also tell us:
- Borland is doomed because they could not produce a product people wanted to buy.
  - Kylix was such a bad name no one knew what it was.
  - Linux developers are as picky as Windows Developers when it comes to paying for software
  - Delphi on Linux was never going to happen and there are better C++ IDEs
  - Borland made another bone head business move
  - Anyone can make sweeping and irrelvant statements base on a single event.

It is time to face reality.  Linux is not going away.  As an MS developer though, I do not see my life ending as Linux expands.  I may need to learn new skills, but guess what, I have to when MS releases their next version of anything.  (Ask the VB to VB.net people)

MSHack
Friday, November 07, 2003

> There have been no spaceships flying to the moon in the last three decades.

What about the Lunar Prospector?

http://antwrp.gsfc.nasa.gov/apod/ap980306.html

Dennis Atkins
Friday, November 07, 2003

Regarding Kylik, I agree with the original poster.

"If only a big name development company would make great development tools for Linux wue would all buy it" the masses shouted. So Borland (you expected microsoft to do it maybe?) did just that. Result -- no sales.

What does that tell you?

Dennis Atkins
Friday, November 07, 2003

How well does Trolltech (QT) do? I have no idea, but I think they sell primarily for Linux UI development.

NC
Friday, November 07, 2003

Dennis, bad logic.

Maybe it occured to you that the 'community' didn't think Kylix was that great of a tool?

Agree with your conclusion but not how you got there
Friday, November 07, 2003

>> What does that tell you?

That Kylix was overpriced at $1500 and has way too many bugs.

The 'masses' didn't say "give us an overpriced and buggy tool and we'll buy it." Give them some credit for having common sense and not saying "It's complete crap but we'll buy it anyways."

Kylix only ran well on one version of Red Hat Linux, and was only available for x86 machines. It is hard/impossible to install and run on Suse, Debian, or any of the other distributions, including current versions of Red Hat.

RocketJeff
Friday, November 07, 2003

I'd venture a guess that Borland's penetration among Linux commercial developers was about as healthy, percentage-wise, as their penetration among commercial Windows developers: they're not doing so hot there either.

Brian McGroarty
Friday, November 07, 2003

The last time I had to do Linux development, I used a product called "KDevelop".  Don't know if it's still around or not but it reminded me of visual studio & I was up & running with it, and being productive, in minutes.  And it was free.

windows hacker
Friday, November 07, 2003

Aww, c'mon ... I used KDevelop, too ... it's no match for Delphi!

John
Friday, November 07, 2003

Hm, you guys could be right. Given the price, one would expect miracles, but if there was stuff that was already better for free, then there's no point. Also, given the number of distributions', combined with the lower marketshare than windows, it was that much harder to break even while maintaining the program costs that much more.

I guess the lesson is that there's little incentive to create commercial software for Linux when there's everything you could want of decent quality for free.

Dennis Atkins
Friday, November 07, 2003

Well John, I should add that I spend about 95% of my time using C++, and the layout is very similar to visual studio, so it wasn't a big stretch.

Dennis, I think you're right:  Who wants to pay for an app when there's quality free versions available.

That's why I hate the open source movement, yet love CodeProject.

windows hacker
Friday, November 07, 2003

Well, "windows hacker", I work with both Delphi and Visual C++.

Delphi is heaps more productive than Visual C++.

Also, it doesn't suck like VB - you can do anything, call Windows API, etc.

John
Friday, November 07, 2003

I mean, in Delphi I can call Windows API without having to import the function yourself from a Windows DLL.

There is simply no comparison between Delphi's productivity and Visual C++'s productivity.

John
Friday, November 07, 2003

Cool.  Language war troll.  It's sooo usenet of you.

windows hacker
Friday, November 07, 2003

Is Novell doomed with Linux ?

Sunish
Friday, November 07, 2003

I don't care if my message is a "troll" or not.

I care if my message is true or not.

And, as far as I know, it is very true.

Now, go ahead and interpret it as a "troll" - I don't mind.

John
Friday, November 07, 2003

Commercial applications cannot succeed on Linux because of the cool free tools available?

Every decent Linux tool get ported to Windows, so how come theres still a thriving market there?

I remember getting junk mail from Borland declaring Kylix as 'Visual Basic for Linux.'  Kylix failed because it was a poor quality product, plain and simple.

Ged Byrne
Friday, November 07, 2003

Kylix was a POS and DOA.  I am sure there is some area for commercial apps to florish on Linux.  Development tools in general, and Kylix in particular, is not one of them.

A pox on the house of Borland for (1) pricing Kylix so high, (2) not updating it as needed to fix bugs and run on the mainstream Linux distros that have been released since the last release of Kylix, and (3) being so out of touch that items #1 and #2 were allowed to happen.  What a bunch of complete dopes.

Want a Borland wake-up call?  Just listen to the archived quarterly earnings teleconferences that Borland management hold for Wall Street analysts.  Borland has been losing their ass the last several quarters, and the Borland talking heads don't seem to have a clue - blame this guy, blame that guy, ... HEY GUYS - FIX YOUR STUFF AND WE'LL BUY IT AGAIN.  The suits don't get this of course.

You think the Kylix fiasco was a stink?  Just wait - JUST YOU WAIT - unitil Delphi .NET (aka Delphi 8) hits the street.  Turds in the punchbowl all around.  You heard it here first.

Mitch & Murray (from downtown)
Friday, November 07, 2003

"It is hard/impossible to install and run on Suse, Debian, or any of the other distributions, including current versions of Red Hat. " - RocketJeff

Could this explain why Linux development will never take off? If we have to make a seperate version of our apps for each and every linux distro then Linux is in a very sad state. - This may be fine for the corporate developers that can dictate (or know in advance) what distribution their users will be using, but for ISVs like my company this isn't possible.

(NB: I have only installed linux 2 or 3 times and not used it extensively, so I am basing this purely on RocketJeffs post and the fact that I respect Borland as a fairly capable company and if they failed in this "cross-distribution" development, then that doesn't speak to well for the chances for smaller companies to succeed in the same field.)

ChrisO
Saturday, November 08, 2003

Yeah, Kylix sucked, but commercial application development on Linux is still stillborn and this trend shows no signs of changing. 

Mister Fancypants
Saturday, November 08, 2003

We thought about distributing a Linux version of our application, but the amount of work required to get it out the door and support it thereafter was prohibitive.  About the only way to have a product that would easily install & run on all the different brands was to distribute the source and have people use autoconf & make - something we weren't willing to do.

windows hacker
Saturday, November 08, 2003

On the binary incompatibilities of various Linuxes...  It seems to me that the Linux culture is based upon everything that you install on a Linux system being in 'make'able packages and therefore, almost nobody installs binary images onto Linux of software that they didn't build themselves.

Therefore, there has never been the discipline in the Linux camp of having to conform to binary compatibility.  You build the tarball and it compiles and inhales in all the library references it needs for the Linux that is building it.

This also means, of course, that "all" software in Linux land is "supposed" to be distributed as source code. So I wonder if Borland, as well as anyone else trying to resell Linux software, ever considered any of this as problematic?

Bored Bystander
Saturday, November 08, 2003

Even distributing as source doesn't cut it sometimes.

Miguel de Icaza of the Mono project relived some of the compatibility horrors they've had with Mono Evolution, the installer for which downloads the source and builds itself (as I recall). Even so, they had a lot of compatibility problems from distro to distro, and even from version to version of the same distro.

They are devoting significant effort to both API and ABI compatibility with Mono. That fact alone may make it the most attractive commercial development environment for x86 and PPC Unix-alikes.

Brad Wilson (dotnetguy.techieswithcats.com)
Saturday, November 08, 2003

First I'll say that I'm a huge Borland supporter, and also make significant use of free UNIX.  In fact, most of my income comes from working in those two areas.

All of the free UNIX implementations, not just the various Linux distributions, are screwing the pooch as far as traditional desktop applications go.  Desktop apps rely on a windowing system of some sort to be competitive.  Herin lies the Achilles heel of all free UNIX.  There's lots of choices for window widget libraries.  Each is a massively complex pain in the tail to install, and all of them are broken in some way. Upgrading is nearly impossible; it's easier and a lot more sensible to wipe out everything associated with X windows and make a clean install.  Even then, it will be hard.  I know because I've done it a few times.

So with this very non-functional model for doing a graphical user interface, anybody hoping to develop commercial, binary distribution desktop applications is pretty well screwed.  As a previous poster mentioned, even doing source distributions is hard, partially because source code compatibility between versions is pretty scarce.

This doesn't mean that free UNIX is a no-win situation.  AbiWord, for instance, is a magnificent desktop application, as is The Gimp. Using a web interface for the GUI is also a really nice way to go, when the application will let you (which frequently it won't).

So I don't predict any great revolution in desktop applications, and I'm not suprised that Borland failed. It was a noble effort though, and I'll continue to use their development tools whenever I can.

Clay Dowling
Saturday, November 08, 2003

"Kylix only ran well on one version of Red Hat Linux, and was only available for x86 machines. It is hard/impossible to install and run on Suse, Debian, or any of the other distributions, including current versions of Red Hat."

Wasn't one of the original goals to put an end to the fragmentation that made UNIX such a nightmare?

Ron Porter
Saturday, November 08, 2003

Gee, I started reading this thread, hoping to find out why Kyle did not seem to fly.

There seems to be two reasons:

1) not very good product
2) not available, or consistent for each Linux dist.

I can’t ever say that as a general rule I heard people bashing, or stating that Kylix was not good. (perhaps only MS haters are vocal!). Is the above “not very good” the real reason here? I mean, we really do need a good RAD tool for Linux. The market is quite good at sorting this stuff out, and surely there is a need for a good rad tool in Linux. Perhps the desktop develpment market for Linux really does not exist.

I have been meaning to look for a good cross platform for windows/Linix. It would naturally seem that Kylix is the ticket here...what the heck happened?

I was also under the impression the Delphi=Kylix. The only difference being the platform. That being the case, then why not develop with Delphi, and then simply move the application to Kylix and re-compile? Was this not one the ideas, or promises of using Kylix?). Again, what the heck happed here?


Albert D. Kallal
Edmonton, Alberta Canada
kallal@msn.com
http://www.attcanada.net/~kallal.msn

Albert D. Kallal
Saturday, November 08, 2003

Commercial Linux vendors have been taking steps to standardize their distributions, so I think the whole distribution compatibility matter is becoming a thing of the past.  I've used VMware across Red Hat 8, 9, SuSE 8, and Debian "sid" without issue, so I have hope things are being sorted out.

Biggest compatibility concerns (IME) seem to be:

-Libc version
-DE/GUI toolkit API and version (KDE/QT, Gnome/GTK, etc)
-Kernel version (if modules are required).  The kernel lacks a consistant device driver API.
-Package distribution: associated dependency issues when building them.  If you're smart about it, you can avoid this pitfall.
-If you distribute source, you will come to realize that autotools aren't so great.  The GNU autotools either need to be massively reworked or replaced with something better.

Jim Battin
Saturday, November 08, 2003

"Wasn't one of the original goals to put an end to the fragmentation that made UNIX such a nightmare?"

No. Linus' original goal was to re-create Minix, but with source available.

Brad Wilson (dotnetguy.techieswithcats.com)
Sunday, November 09, 2003

The problem with widgets, drivers, etc., combined with the X-Windows stupidity that seems embedded in Linux forever, is the major impediment to desktop acceptance.

Compared to a Windows environment on an equivalent configuration, X SUCKS. And I don't want to tell you what I *really* think :)

You won't get good, commercial RAD until there is a stable and desirable desktop environment for its development.

This was the core of the Kylix problem. It felt like one of those PC products ported to Macintosh, which retained the PC feel. It crashed too often, and wasn't able to be run as the various (fragmented!!) portions of the operating environment were upgraded.

Borland deserve credit for trying, but condemnation for not trying hard enough. Linux also deserves blame for being an unco-ordinated pile of crap masquerading as a competent desktop environment.

HeWhoMustBeConfused
Sunday, November 09, 2003

Borland made two mistakes:

1. Name Delphi with unpronounceable K*!%x. Yeah... it must be similar (but it's just similar, nothing more) to Linux.

2. Instead of bringing Delphi for Mac OS X, they did it for Linux. They could've done it for Linux too, it would be nice-to-have.

But, it seems Borland marketing was making decisions by reading way too much slashdot where the posters (mostly) have not the faintest idea of real world problems and have enough time to wage a war for or against Linux (or Microsoft for that matter).

Enjoy Cocaine
Sunday, November 09, 2003

"If only a big name development company would make great development tools for Linux wue would all buy it" the masses shouted. So Borland (you expected microsoft to do it maybe?) did just that. Result -- no sales.

"What does that tell you?"

It tells me that marketing departments are entirely composed of morons. It also tells me that anyone who thinks that the performance of a business plan based on that idea is an accurate guide to the performance of business plans based on accurate information is equally intelligent.

Here's a more accurate version of that statement:

"If only a big name development company would make great development tools (where 'great' includes 'does things we simply can't do for free already') for Linux some of us would buy it" the masses actually thought.

Furthermore, real sales were likely to be to professional developers, rather than amateurs, who will have significant incentive to not change langauge and development environment at one time, and thus would be likely to pay for a quality GUI designer and IDE that uses their existing compiler, but would not pay for an entirely new language and development environment. An IDE that requires the use of nonstandard extensions to the compiler, or a compiler for a proprietary language, is unlikely to succeed.

The most likely people to pay for a commercial development environment were the ones who, due to the choice of language, were the least likely to actually pay for Kylix. This isn't because the language is bad, but because it was new to Linux - and existing Linux developers don't use Delphi.

Porting Delphi to Linux was doomed from the start. Porting the IDE to Linux and adapting it to use the gcc compiler could have worked, and would have provided a base from which to develop a Delphi compiler that could use that IDE.

andrew m
Sunday, November 09, 2003

"in Delphi I can call Windows API without having to import the function yourself from a Windows DLL"

On Linux that would be a pretty clever thing to do.


Monday, November 10, 2003

Looks like the solution will need to come from Gnome and KDE, and developers who write outside the documented APIs for these desktops will need to be eliminated by the marketplace.  If there's a critical mass of "pure" Gnome and KDE apps, at some point X could be swapped for some other graphics layer, as long as all GUI calls were properly abstracted.

Jim Rankin
Monday, November 10, 2003

So what of the new C++BuilderX product?  Borland is touting it as a multiplatform development tool that leverages existing C++ compilers in a new, comprehensive environment.  Could this be Borland's second approach to the Linux market, and this time the right one?

JT
Thursday, November 13, 2003

"Could this explain why Linux development will never take off? If we have to make a seperate version of our apps for each and every linux distro then Linux is in a very sad state."

Sun StarOffice seems to do fairly well being distributed in binary form and works fairly well on everything I've tried.

Nicholas Smith
Tuesday, March 30, 2004

*  Recent Topics

*  Fog Creek Home