Posted Apr 8, 2005 2:30 UTC (Fri) by ncm
In reply to: A shame
Parent article: GCJ - past, present, and future
Should I make a list? The language's purpose was not a way to express programs clearly and efficiently, but rather to be a weapon for Sun to use against Microsoft. It's rife with peculiar design decisions that make no sense except in support of proprietary head-knocking. Free Software doesn't need any of that; we already had "write once, run anywhere" that actually works. We don't need to give up anything to get it. The language has no public standard, and its originator has no interest in making one. That makes it unsafe as a development platform for careful people; the future of the language is not controlled by people who use it.
Besides peculiar choices that were more or less purposeful, the language and its libraries were designed in enormous haste by people who really didn't understand quite what they were doing. Therefore, many of the central choices were simply stupid. Examples of that include making member functions virtual by default; compulsory exception specifications; a character type 16 bits wide, not wide enough to represent a whole character, but too wide to avoid endianness problems; syntax that mimics the historical mistakes of C for no plausible reason; no way for user defined types to mimic built-in types. Because no significant program can avoid pervasive casting, it suffers all the inconveniences of a type-safe language without actually providing type-safety. Finally, it is saddled with the academic "garbage-collection" disease, which is one reason why real programs are so slow, despite that benchmarks suggest it should be otherwise. Sophisticated techniques to fix that have been promised to come out any day now for the last ten years. (C# doesn't fix many of its problems, even given years of hindsight.)
The language is a snapshot of how "object oriented" programming was understood (or misunderstood) toward the end of the overhyped '80s. That's hard to excuse considering it came out half a decade later, and it is entirely ridiculous fifteen years on. Object-orientedness has turned out to be useful for some things, and irrelevant for most; while it's handy for a language to support it directly, it's foolish to make every detail of a language depend on the notion.
Design concerns aside, it's notoriously hard to deploy Java programs, because of the JVM incompatibility problem. Each program typically either requires the default platform JVM, if there is one, or ships with its own. A system with six Java programs typically has six JVMs installed.
Fortunately we have a public-standard language that is more powerful and more thoughtfully designed than Java, and you have literally hundreds of end-user Free Software programs written in it on your disk right now. It's unencumbered with submarine patents, fast and efficient, and has lots of fantastically powerful libraries available. It works well enough that you never need to know -- and generally don't know -- that a program was written in it. Can that be said about any Java program?
to post comments)