Slashdot Mirror
Borland Divests IDEs to Focus on ALM
ShinyBrowncoat writes "Borland recently announced they are putting their IDE business up for sale (JBuilder, Delphi, etc.)." This move comes at the same time Borland announced they would be aggressively pushing forward with their Application Lifecycle Management (ALM) business by purchasing Segue Software Inc.
So anybody want to start a collection to open source them?
Nothing that I've found yet seems to be a good stepping-stone to migrate away from oodles of code written in C++ Kylix. Lazarus is great if you wrote your Kylix apps in Pascal, but for people tied to C++, it's not a happy situation. We finally opted for a split solution in the form of KDevelop+QTDesigner for some stuff, and to C#+GTK# for some other stuff, but I'd love to hear what other peoples migration experiences have been. None of these free IDEs offer quite what Kylix did in terms of ease-of-use, RAD and cross-platformability.
They ignored the Open Source movement (Linux included) and hobby programmers. After Microsoft had beat them with Visual Basic, they could only keep their loyal market. What's after that? Death.
They should have followed Netscape's example and opensource their IDEs.
But no - instead, they decided to overinflate their prices, and well, the rest is history (pun intended).
Online Petition for Mark Shuttleworth, so we all can enjoy, port and improve the Borland programming tools
I absolutely loved the early versions of Delphi. The manuals that came with it were long, involved, and brilliant. It was like being taken on a tour of what programming should really be like by about ten of the smartest guys in the business. Writing Object Pascal felt, much of the time, like writing poetry. The component library was clean and beautifully laid out. The IDE was super-responsive. And it could compile code faster than anything on the planet at the time. Back in the days of the 486, compile time really mattered, and being able to do 10,000 lines per minute on a 486-33 was extremely impressive. (hopefully I'm remembering my numbers correctly, it HAS been a very long time... it might have even been 100,000, but that seems too fast for a 486. Whatever the actual number was, it was, god, twenty times faster than anything else.) And a compiled Delphi program was just one EXE. No DLLs, no runtime, no dependencies, no distribution headaches... one EXE you could dump on a floppy and hand to someone. And the code was lightning-quick.
But then it started going in a strange direction... after Delphi 3, they decided to focus totally on database programming, and they ignored most of the other good stuff. And somewhere in that time frame, Microsoft swooped in and bought Anders Hejlsberg, the real brain behind Delphi. They correctly identified him as THE guy at Borland, and paid him a cool million in hard cash, upfront, to come to work for them. We are seeing the final results of losing Anders now. Without him at the technical helm, Borland entered into a long, slow decline. Delphi went off the rails, they forgot what was really great about it... it turned into a bloated mass of crud, focused on a tiny subset of the full universe of programming.
And then there was Kylix, which was an abortion if I ever saw one... what a horrible piece of software. I coughed up $1200 for the first Pro version because I was excited to see Delphi on Linux.... except it really wasn't. It looked like Delphi, but it didn't feel like it. It was still fundamentally a Windows program, with the minimum amount of effort needed to port things. Distributing a Kylix app was freaking impossible if you didn't already understand the Linux library system very intimately. There was nothing at all like the 'single-exe' feature, even though they made claims about 'easy distribution' on the box. And the documentation was terrible, just incredibly bad.
Seeing Borland die at this point would be more of a relief than anything; they have become a clueless company and haven't got a prayer of long-term survival. They have pissed all over everything they've ever done. You'd have to be an idiot to choose their software these days, between the freeware and the commercial alternatives.
For Microsoft, hiring Anders was a brilliant move; destroy a competitor for just one million dollars, pocket change from their standpoint. Anders worked on language recognition for awhile, but eventually he went back into compiler technology. He's the main brain behind this little language you might have heard of, C#.....
I don't know in the US, or in other countries, but here in Brazil Borland is still relatively popular. For a long time, Borland had the lead on development tools. But since it started to fall apart, it never recovered, and it's just a shadow of what it once was. Many people blame Philippe Khan, others blame the subsequent CEOs and the whole Inprise imbroglio. But I prefer to look at it from a programmers perspective.
I started Turbo Pascal 2.0, on floppies. I remember seeing the ads on Byte Magazine. For anyone who tried Pascal on CP/M, or in USCD's implementation, it was a dream come true. And it was really fast! Later, I worked with all versions - from 3.0 to 5.5, and then Borland Pascal 6.0, with object orientation and Turbo Vision, a character based event-driven framework. I have the impression that Borland at that time tried too much, too hard; they tried to change paradigms, to change the way we programmed, but it was too big a change at once. But history does not stop here. Borland managed to get a lot of things wrong in a couple of years. Quattro was ok, but lacked the 'extra something' that made Borland special. Paradox was innovative for its time, but its stability was never something to write home about (IMHO, it managed to be worse in this respect than Access, and I'm giving my personal testimony on this). Borland even tried to run the clock backwards and sell a text processor named Sprint that I'm sure only the true dinossaurs around here will remember hearing about.
However, Borland still had some gas, and a new chance to get things right. A few years later, I got my hands on the Delphi 1 beta - it was a eighteen 1.44 floppy install, in a time when CDs were still far from popular. The quality of Delphi was amazing - they just got it right. But by then, VB had a small edge. For some reason, and for lots of small misteps, Borland gradually started to lose the lead.
I still can't get what happened around the whole Inprise situation. That they opensourced Interbase, just to close the source later, is something that I don't understand. They also got the pricing wrong. Borland always had the lead on low cost tools, but it started to charge one arm and one leg for a usable toolkit. The 'personal' editions were crippled, and missed some features that almost everyone needed (such as compiling ActiveX controls, or using the database controls in the library). It started to lose touch with the developers. The community (a vibrant one) started to look for other tools, just at the time when open source was starting to become mainstream.
By the way, even in the pre-Internet days, the community was amazing. One of the first popular software repositories in the Internet was Professor Timo Salmi's ftp.uwasa.fi. There were huge repositories of Pascal componentes, many of them in eastern Europe - Poland and Russia, for example. Borland could have amassed the power of the community, but for some reason, it largely ignored them. Students, once one of the strongholds of Borland penetration, were also ignored.
It's a shame that a company like Borland had to go this way. I personally would prefer that the ALM division was divested with a new name, so that Borland, the company, could be allowed to die with dignity. Perhaps a new structure - a Borland Foundation perhaps (borland.org anyone) - could pick the bones to start again. But I fear that's too late, even for that.
Later, with Borland C++ 3.0, "a = 1;" worked, but I could never get programs that used more than 64K to work right because the compiler couldn't handle segments correctly. I didn't realize what was really going on, just noticed programs doing weird stuff, fiddle with something that wasn't related to the real problem and then see the program mysteriously work right.
Having written dozens of professional applications with Borland C in my time, I have to say that this is simply not true. Borland C (BC 3.1 in particular) was a very robust C/C++ compiler, even though it didn't have the best optimizer (compared to say, Watcom). The problems you saw were most likely caused by small bugs in your applications - stuff like forgetting to initialize variables, ignoring the 64K segment limit, not using "huge" pointers, etc.
One of the first things I learned as a programmer was never to blame my problems on the compiler - it isn't very productive, besides the compiler cannot defend itself :-)
Specifically, US patent 5,628,016 on structured exception handling. This patent is preventing the Wine, ReactOS, GCC and MingW people from supporting exception handling that is compatible with the Microsoft implementation.
You are right!! I guess I must eat my hat now :-)
Since I like to fiddle with compilers and I have many santimental memories of BC 3.1, I downloaded pdcurses, fixed the DOS makefile (which was broken, by the way!) and reproduced exactly the problem you are describing. After half an hour of digging, I found out that the bug is not caused by register variables, but by induction variable optimization. It is illustrated by this sample code:
BC synthesizes a pointer to track the value of &buf[y], and an increment value sizeof(buf[0]). The pointer is updated with the value on every iteration to avoid an address calculation. When yd turns negative, the increment value is supposed to be set to -sizeof(buf[0]), but it isn't. This is pretty bad :-) I could post the assembler listing, but that would be too much ...
Anyway, thanks for spoiling my rosy memories :-)