Actually he doesn't....he considers Clang/LLVM to be open source and not free, Clang/LLVM does not restrict users from developing their own extensions and licensing them as they see fit which ironically makes the software not free (according to the RMS ideological software freedom purity test).
So when RMS says that, he's saying Clang/LLVM opened up the source code in order help their software perform better and see wider usage. Whereas RMS's goal for GCC is to ensure freedom for its users. Technical superiority is great, but it is not the goal, and may be sacrificed if necessary to achieve that goal.
He's not saying Clang isn't free software, he's saying their ultimate goal is not freedom, as contrasted with the FSF.
1) Free means freedom in the copyleft sense.
2) Free means freedom in the open source sense.
3) Free means free as in beer.
In the article RMS appears to be using definition 1, elsewhere on the FSF site the FSF uses 2.
"In the free software movement, we campaign for the freedom of the users of computing. The values of free software are fundamentally different from the values of open source, which make "better code" the ultimate goal."
It's very clear that "free software" in the article is definition 1, since MIT/BSD falls clearly in category 2, LLVM/Clang is NOT free software according to the definitions used in the article, and therefore the position RMS is advocating in the article is that LLVM/Clang is not free software, but rather open source software.
Yes to everyone else in the world, including the FSF, Clang/LLVM is free software but according to RMS it is not.
It is possible to be both unfree, and support dictatorships, like the USSR. RMS's goal with GCC, is to be like the mythical free country that also supports freedom and discourages dictatorships elsewhere in the world.
In other words, your software can be free, and yet support and aid those who would take your freedom away. This is the claim RMS is making of Clang.
"The values of free software are fundamentally different from the values of open source, which make "better code" the ultimate goal."
Perhaps you could provide quotes where RMS in the article says that clang is 'free software'. According to RMS in the article clang is a major set back for free software. (because it's better than their piece of crap)
http://www.gnu.org/licenses/license-list.html
clang is "free software" according to his own definition. It is entirely possible for a specific free software program to be a setback to the goals of the free software movement. His position is that copyleft licenses (very often, and certainly in this specific case) advance the goal of freedom more effectively than non-copyleft licenses. The differing values in this case basically amounts to whether one considers it worth using a copyleft license to ensure that the software being used remains free, or whether one considers it a drawback that a copyleft license may prevent it being used in proprietary software.
Furthermore which compiler-suite served as so long so good in the "dark" pre-LLVM ages?
Biting the hand, which feeds you comes to mind...
If the FSF consists solely of the opinions of RMS then I retract the previous statement and would say that the page mentioned expresses an opinion inconsistent with what is expressed in the email.
"The values of free software are fundamentally different from the values of open source, which make "better code" the ultimate goal."
Clang wants to make a great compiler, they chose the NCSA license to facilitate that, according to the article makes it open source software and not 'free software' rather than 'unfree software'
I get you now, makes perfect sense.
This restricts users freedom, that is his point.
GPL is not about the rights of the programmer, it is about the rights of the end user.