Interesting take on the OS front.
http://sta.li/
http://www.infoworld.com/article/30...ribution-smashes-assumptions-about-linux.html
http://sta.li/
http://www.infoworld.com/article/30...ribution-smashes-assumptions-about-linux.html
I somehow have serious doubts about that claimThe most obvious benefit is that static binaries have a smaller memory and on-disk footprint.
Not entirely true. The page sharing only works if your code is PIC, which a lot of the shared libs seem not to be. So once you need to patch offsets in the code, your page can no longer be shared. Things like errno come to mind here. But I have fought this fight with some GNU die-hards, who would not care to argue on a fact basis about this.Quoting the linked article:
I somehow have serious doubts about that claim
edit: It might be true for the on-disk part when your whole system only uses a handful of binaries ... but as soon as several binaries start sharing more than just libc.so (or the number of binaries gets bigger, so that all these copies of libc.so add up to more than the shared linker infrastructure), this is just ridiculous. Not even talking about memory footprint, see page sharing etc ...
Obviously true, but so far I thought position independent code was standard for shared libs. At least, I always compile my own withThe page sharing only works if your code is PIC
-fPIC
.It should, as far as I have understood the virtual memory system. And I must have read almost all of itDoes FreeBSD actually support page sharing with shared libraries? I have a vague memory about reading on the mailing lists that it's still unimplemented.