4 ms·

Leadership failure? How is your compiler doing? By your argument any second place product is a failure. Apple didn’t develop LLVM. It was developed at Unive

Leadership failure? How is your compiler doing? By your argument any second place product is a failure. Apple didn’t develop LLVM. It was developed at University of Illinois at Urbana-Champaign. Yet again, Apple had taken something invented somewhere else and made a killing off it. RMS has a strong political stance and the result is software that has benefited billions in one way or another. Lack of faults is the worst measure of strength, but it’s cheap for people who have done nothing to throw shade and feel superior.

geofft 6y ago
> Leadership failure? How is your compiler doing? I'm enjoying rustc very much, thank you! I hope that gcc-rs will catch up soon, but until then, I'm thrilled that the Linux kernel compiles just fine with Clang and there are millions of people running Clang-built kernels, because it means adding rustc into the mix is straightforward. I'm also enjoying using libLLVM in projects from Numba to llvmpipe. Numba is helping us speed up some serious research computations at my current job, and llvmpipe solved a major practical problem with GNU/Linux on the desktop at a previous job. Those projects could not have been built on top of GCC. I'm also enjoying all sorts of analysis tools that use libLLVM and libclang (include-what-you-use being a very simple example) even on code that I compile with GCC for legacy reasons. If you look at the linked threads, RMS wants such tools to not exist, or at least to be hard to write and buggy. > Apple didn’t develop LLVM. It was developed at University of Illinois at Urbana-Champaign. Not disputing this. Apple didn't develop GCC either. They contributed to both. Stallman was worried that if he made GCC either permissively-licensed or modular, companies like Apple would privately fork the compiler and not contribute back. He thought it was a vindication when NeXT contributed back their Objective-C frontend, which they were going to keep private until he threatened them with legal action. But UIUC made LLVM both permissively-licensed and modular, and Apple contributes back to it nonetheless, with not even the slightest legal obligation to do so, making a killing while also producing a free software compiler that is superior to GCC. Everything GNU tried to do by technically hobbling GCC, UIUC actually did by producing a superior product in LLVM. My whole point is that Apple didn't develop LLVM, but that they contributed back to it because they saw a financial advantage in doing so. > RMS has a strong political stance and the result is software that has benefited billions in one way or another. Correlation is not causation. I'm not claiming that GCC was unsuccessful; I'm claiming that RMS's strong political stance did not make it so. And RMS would be the first to argue that mere popularity is not the sign of a software product's success. GCC is supposed to protect a user's freedoms more strongly than LLVM, yet a user is practically more free with LLVM; for instance, I am free to get the AST of my own software. For those of us who believe in the ideals of free software - which includes both me and RMS - this is a sign of a strategy that failed, somewhere, that should prompt some reflection.

geofft 6y ago

> Leadership failure? How is your compiler doing? I'm enjoying rustc very much, thank you! I hope that gcc-rs will catch up soon, but until then, I'm thrilled that the Linux kernel compiles just fine with Clang and there are millions of people running Clang-built kernels, because it means adding rustc into the mix is straightforward. I'm also enjoying using libLLVM in projects from Numba to llvmpipe. Numba is helping us speed up some serious research computations at my current job, and llvmpipe solved a major practical problem with GNU/Linux on the desktop at a previous job. Those projects could not have been built on top of GCC. I'm also enjoying all sorts of analysis tools that use libLLVM and libclang (include-what-you-use being a very simple example) even on code that I compile with GCC for legacy reasons. If you look at the linked threads, RMS wants such tools to not exist, or at least to be hard to write and buggy. > Apple didn’t develop LLVM. It was developed at University of Illinois at Urbana-Champaign. Not disputing this. Apple didn't develop GCC either. They contributed to both. Stallman was worried that if he made GCC either permissively-licensed or modular, companies like Apple would privately fork the compiler and not contribute back. He thought it was a vindication when NeXT contributed back their Objective-C frontend, which they were going to keep private until he threatened them with legal action. But UIUC made LLVM both permissively-licensed and modular, and Apple contributes back to it nonetheless, with not even the slightest legal obligation to do so, making a killing while also producing a free software compiler that is superior to GCC. Everything GNU tried to do by technically hobbling GCC, UIUC actually did by producing a superior product in LLVM. My whole point is that Apple didn't develop LLVM, but that they contributed back to it because they saw a financial advantage in doing so. > RMS has a strong political stance and the result is software that has benefited billions in one way or another. Correlation is not causation. I'm not claiming that GCC was unsuccessful; I'm claiming that RMS's strong political stance did not make it so. And RMS would be the first to argue that mere popularity is not the sign of a software product's success. GCC is supposed to protect a user's freedoms more strongly than LLVM, yet a user is practically more free with LLVM; for instance, I am free to get the AST of my own software. For those of us who believe in the ideals of free software - which includes both me and RMS - this is a sign of a strategy that failed, somewhere, that should prompt some reflection.