I make the majority of my income from Linux. My longest job was at large Linux company, who makes most of its revenue replacing Sun hardware.
There is very little I like about Solaris. Ancient tools, badly named packages, a messed-up filesystem that puts binaries in /var, very few supported apps on x86, and ZFS, which solves the mid 1990s pre-san problem of having massive amounts of data connected to individual hosts, rather than use a SAN filesystem, like, er, everyone.
DTrace, however, shits all over SystemTap, for the following reason:
* I can understand DTrace
* I cannot understand SystemTap
This might be a massive problem with SystemTap documentation, but I doubt it - RHEL usually has better doco than Sun does.
DTrace is the result of a few people dedicated to making it work well. It has been around for a few years and has had time to mature. I'm so glad to have it on OSX. Even so, it's a little irritating that to make good use of it you need to dig through a lot of Solaris stuff.
From reading that kernel summit mailing list thread I get the impression that the Linux folks only think that dtrace is useful for system administrators and kernel developers.
What they fail to realize is that by supporting dtrace they will make Linux a better application development platform too.
I've been in many situations where I am developing code on Linux (in C or C++) and where I need to understand what is happening in my code or the system. The current tools on Linux are either completely underpowered or non-existent, so I frequently move code to my OS X box or even to OpenSolaris running in VMWare. Just to use dtrace.
I honestly can't see what's in it for Sun here. They have an excellent, free kernel of their own in OpenSolaris. Dtrace is one of Solaris crown jewels. What possible advantage could it be to them to invest in two different free Unixes?
There is very little I like about Solaris. Ancient tools, badly named packages, a messed-up filesystem that puts binaries in /var, very few supported apps on x86, and ZFS, which solves the mid 1990s pre-san problem of having massive amounts of data connected to individual hosts, rather than use a SAN filesystem, like, er, everyone.
DTrace, however, shits all over SystemTap, for the following reason: * I can understand DTrace * I cannot understand SystemTap
This might be a massive problem with SystemTap documentation, but I doubt it - RHEL usually has better doco than Sun does.