After reading carefully *all* replies in slashdot, phoronix and the list, I wanted to reply to a few topics that have shown recurrently, and are not fitting in the GCC list.

1. RMS should be worried

No, he shouldn't. This is not about mixing licenses, it's not about taking away freedoms and it's not about stealing GCC's shine. RMS's contribution to society cannot be overstated, and I don't mean to obfuscate the importance of GPL, GNU, etc. This simply has *nothing* to do with politics, or copyright, or patents.

2. Only LLVM will benefit, because GCC can already use LLVM's code

While the latter is true, it doesn't imply the former. Also, this is not about being better than GCC, it's about both toolchains being better to the users, which I'm am both. This is not about competition, but collaboration.

Some people say GCC is going to die soon, I disagree. Other people say GCC will rust a bit with all the new blood going to LLVM, that might be a bit more real, but still, highly exaggerated. In any case, GCC is not immune to the outside world. With LLVM being actively encouraged by the kernel community to be compatible, the "one true compiler" position is being slowly replaced by a "number of free/open toolchains available", and in that scenario, GCC will benefit from collaboration as much as LLVM.

I can't read the future, but if you ask me, collaboration is always better, no matter in which position you are.

3. Competition is good for both on innovation

This is true, but collaboration is *also* good. We're talking about free/open software, we can both collaborate where competition hurts our users, as well as compete for performance and new features. I'm not proposing on merging the two toolchains, that would be outright madness! Just that we agree on the size of our nuts and bolts.

4. Enforcing standards & discussions will curb innovation

Absolutely right! Every second we spend arguing is a second we don't spend coding. My idea is to have a sort of zero-cost model, where tools report *how* they do it and maybe even for what reasons, and other tools either agree on, or disagree. A discussion will only happen if there are disparate solutions AND both sides want to argue, which no one should be forced to.

This could wind up in two threads: either every one posts what they think is right and don't discuss anything, or discussion ensues, standards are proposed upstream (C++, ISO, POSIX, Dwarf, etc) and compilers implement a more sane interface and the users benefit. Either way we win, since at least we'll have some documentation. The third outcome is to no one submit anything, than, well, no one spent time anyway, so we haven't lost anything.

5. Other standards should be used instead

Indeed. But standards are slow, and for a reason, and compiler implement extensions that will become standards in the future. This is how it's always been and I don't see this moving away. If most/all free/open compilers implement a specific feature, it'll be more argument to the standard to adopt that feature.

Other bits like warnings, implementation of standard classes, data layout and things that really promote binary compatibility are toolchain specific, and if all agree, than we could *use* a mix of tools interchangeably. This is a win for all the users.

This is exactly what I intended to do. I don't want to create extra work for compiler engineers of other toolchains, as well as extra work for compiler users, to ifdef their code to a level beyond recognition. We already have too many GCC extensions, and creating Clang extensions at this point wouldn't help anyone. If toolchain providers (I'm mostly interested in free/open ones) could agree on some basic levels, users would know what to expect from their compilers without the need for magic headers.

It's still possible to have a game right-priced, even if it's worthless to you.

If I buy 1 game a year is too much, but when I do, I don't mind spending $50 or even $100 if I feel the price is right. I still play them once or twice, but the price was right.

Take X-Plane. I've paid more than $100 including taxes and shipping for the version 9. I've played a few times and now it's somewhere in my HD. But the software is amazingly well done, no stupid root-kits or DRM, and I could easily get that via torrent. I just didn't.

We shouldn't pay to *have* the media (pointless these days), we should pay to *please* the authors. Be it with games, music, films or whatever art.

Hum, so Formula 1 cars should cost $10. Not because they're drivers are cheap, but because they only run one day month and cannot go outside the circuit

There is no such word in bioinformatics. Been there, did that, all failed. Everything is a hack and the high-level has no idea of what software or software quality is at all. All of them are scientists, none is technical or developer and barely anyone really knows what software is. Hal: there is no such thing as VP of software in bioinformatics. There is barely informatics... Thegrassyknowl: briefings works when the audience at least have a clue on what you're talking about. Bioinformatics is doomed!