A lot of people are dissing DTrace, saying “Bah, Linux has had that for ages” (wrong) or “So, when will it be ported to Linux?” (not any time soon, you putz), simultaneously telling the developers of DTrace they’re stupid, Sun sucks, and clubbing them about the head with Linux. All of which is a stupendous failure to recognize the utility of this thing, in the context it was intended to be used, at least initially. Yes, there’s the occasional “Yes, this is good stuff” line but it’s almost always dwarved by the following “but it’d be so much better in Linux” paragraphs of crapola. Do these people think Mr. Torvalds would approve of their behavior? I don’t.
I’ve been using Linux in daily life since version 0.9.1 (circa 1993/4 I think) so obviously I like it. I make my living as a system admin for Linux and Solaris. I’m also no fan of many of Sun’s business decisions over the last several years; but, I’m just geek enough not to give a crap about that as long as they’re producing and making available good technology of some kind and not being unacceptably evil in the process. ZFS, containers, and DTrace, just to repeat the three of several dozen that everyone is focusing on, qualify not just as good technology, but also freakin’ innovation, a fact that no one outside of Sun is really recognizing at all.
Just because something is not already in Linux does not make it invalid, uncool, a waste of time, or whatever other negative you can think to give it.
Linux is good, yes; but it is not the only good, and never will be. Period. Were it not for things like Microsoft, Sun, HP, Novell, etc, to compare Linux to, it would suck just as much as they do because something better would come along in fairly short order. It’s about choice, not dominance or l33t haX0r, or even necessarily about the GPL (although that is likely the very reason Linux is what it is). Microsoft got to where they are because they made the world believe they were the only choice — anyone remember GeoWorks, Microsoft’s competition back in the Win3.1 days? GW was a far superior product but they didn’t have MS’s marketing budget so they’ve been relegated to the embedded OS market. If Linux were the only choice, it’d be the same game with different players.
If Sun were really clever, and they’re decidedly not that, they’d do what companies have done for decades when faced with declining PR value: change the name, either of the company or of the products it sells. Solaris 10 is going to carry, like it or not, the baggage, real or perceived, of all previous versions; and, believe me, there’s plenty of baggage. So, if S10 is such a departure from previous versions, maybe they should have named it something else. In the geek world of version numbers, S10 is nothing more than a minor upgrade because, in the OS itself, its version number is “5.10” — the previous version, Solaris 9, was “5.9”. If it’s so different, make it look different or no one will know they should look more closely at it. It’s not the sysadmins/geeks you’re selling to here as much as everyone else. We in the choir appreciate the sermon, especially when it has so many interesting bits to it, but you really should be preaching to the non-admin/geek audience in terms they can comprehend. If you can do both, all the better; but if you have to pick one, the latter is likely the better choice right now.
Now, for some juicy geekness about DTrace please read the blog this link points to:
The Observation Deck
With my explanation of a demo gone wrong, several people have asked me the more general question: how does one demo DTrace? This question doesn’t have a single answer, especially given that DTrace is best demonstrated with ad hoc queries of the system. Indeed, the best demos are when someone in the audience shouts out their own question that they want to see answered: answering such a question instantly with DTrace nearly always blows away the questioner — who has presumably suffered in the past trying to answer similar questions. Of course, saying “why, there are many ways to demo to DTrace!” is a useless answer to the question of how one demos DTrace; while there are many ways that one can demo DTrace, it’s useful to get a flavor of how a typical demo might go. So with the substantial caveat that this is not the way to demo DTrace, but merely a way, let me walk you through an example demo.