Well that’s what I said back in 2002 — and it certainly had an impact within Sun — but I see Darrin has toned it down a little in his inspirational list of performance mantras. Somehow “if a competing OS is faster, it’s a Solaris bug and it’s important” doesn’t have quite the same punch!
I still stand behind my original call to action. And I’m glad to report that the gap has been closing and continues to close. Of course there will always be cases where the RAS (reliability, availability, and serviceability) or scalability of one platform make it slower than a competitive but less capable platform. However, the days of attributing all performance vices to other mitigating virtues are long gone. Quite often code is slow simply because it is slow and needs tuning.
I’ve already promised to blog about my lockless scalability fix for getenv(3C) (and I will do so soon). This is just one instance of me putting my putback where my slogan is. Although in this case it would have been possible to take the moral high ground since the other platform’s implementation was only faster because it was completely multithread unsafe — it may still be so for all I know, but I do know that Solaris 10 now has a fast and safe implementation.
Is it too cheeky to suggest a corollary? …
“If Solaris has a cool distinguishing feature, it’s a Linux bug!”