Game 10: Running into a brick wall called Halladay

I admit that I didn't get a chance to watch any of the game tonight. Oh, well.

Astacio pitched 7 innings and gave up two runs - very decent, but Halladay looks like he's returning to Cy Young form early on this year. The rest of the league needs to be on the lookout.

Dickey's been placed on the DL and called up Regilio - who pitched two scoreless innings in relief. And, it looks like some pitchers are on their way back. Francisco and Almanzar are reportedly going to be activated for tomorrow game; I'm not sure yet who they'll send down. Regardless, they'll provide some much needed help to the collection of firestarters present. At this point, they could pretty much randomly point to anyone in the bullpen and send them to AAA and it'd be the right call.

[/sports/rangers/2005] permanent link

Attended the OpenSolaris BOF at Usenix: Part II

See Part I

My second comment after listening to the BOF and the Sun employees speak about OpenSolaris is that Sun wants to place themselves as challenging Linux. That makes sense in the marketplace and perhaps even in the broader sense. Yet, where it falls down now is in the technical analogy. I don't believe the Solaris engineers have settled upon a clear and compelling analogy as to how OpenSolaris is different and better than Linux to a developer or a system administrator.

In the BOF, the Sun developers kept tripping over themselves trying to explain how they provide more than just a kernel. The answer should have been: "Everything in the OS is part of OpenSolaris." You can leave the qualifications about legally encumbered binaries and source out of it. People can realize that: Linux has binary-only drivers too. It's not worth obfuscating the conversation over those little details.

In my opinion, the close technical comparision as to what OpenSolaris will offer has to be BSDs - such as FreeBSD, NetBSD, OpenBSD, DragonFly, etc. The BSDs essentially provide the same deliverables that OpenSolaris would when all is said and done - the BSD camp has releases available today and done so for years (heck, decades). Hence, in turn, the arguments that the BSD people often make against Linux may very well apply to presenting the ideal OpenSolaris vision.

For many reasons, I much prefer the governance model and the philosophy behind FreeBSD over Linux. I'm not at all a fan of the 'benevolent dictator' project: Linux just happened to reach a critical mass and was extremely lucky. But, there's no way that I think that could be successfully and consistently copied. In fact, every time I've seen this behavior on other projects, they've been complete disasters. (qmail anyone? sourceforge?)

The FreeBSD folks have a set review process and a release engineering team. Yes, it's not on the scale of Solaris's review processes. However, FreeBSD is just a bunch of unpaid volunteers, too. So, the FreeBSD folks have to make do with what resources they get from the community. It's not realistic for them to make it impossible to make progress: no one wants a project that doesn't release. I know that the Solaris team places a lot of emphasis on code quality and stability. Excellent. Glad to hear it. So, does everyone else on a high-quality project. The difference, I feel, is how much resources we can realistically spend on it. Sun, by virtue of their pocketbook, can place a lot of emphasis on testing: vastly more than a strictly volunteer-based organization.

In comparison, Linus explicitly believes that it is the role of the downstream distributors to deal with quality and testing. Now, that's not necessarily invalid, but it's not the model I prefer. "The buck stops THERE" is a great way to end up with unmaintainable code. The Linux source code I've seen has uneven code quality: some places are great, others very poor. There's little consistency. That's why code reviews are essential!

And, this, of course are two reasons why I think OpenSolaris may indeed be more successful than Linux in the long run: a much stronger governance model and higher-quality code. Yet, that honestly guarantees little.

And, if it survives the initial onslaught of Linux, my question is how is it going to differentiate itself long-term from FreeBSD? Does it even need to do so? Perhaps so, perhaps not. I noticed David O'Brien from FreeBSD in the BOF: I wonder what he thinks of this.

It'll be a very interesting experience to watch this experiment unfold.

[/software/solaris] permanent link

Attended the OpenSolaris BOF at Usenix: Part I

Roy and I decided to attend the Developing for Solaris and OpenSolaris BOF at Usenix '05 tonight up in Anaheim. Roy, of course, is on the CAB. And, I'm part of the OpenSolaris pilot program (representing the ASF infrastructure team).

The BOF started with an introduction to debugging and the tools available for development. (zones are yummy.) They went into some depth on smf. smf, in return for dropping init scripts, gives you a lot of flexibility and power in controlling the init process via XML files. That's nice. Except I'm not sure we (as admins or developers) really needed this: init scripts worked for the last twenty years - it wasn't just begging to be replaced, in my opinion.

Random question: how much slower does the XML-based smf system make Solaris in booting? I'm specifically thinking how much work Thom and the Canonical folks did on speeding up the boot process of Ubuntu. I just can't imagine parsing XML is a world-beater; but frankly neither were shell scripts.

I realize that the console login starting much earlier in the init process is great: but really, that could have been done with init scripts too. Just background the init output into a virtual terminal. If that was the real goal, it could have been done without adding XML. (XML is not the answer for world hunger.)

In sum, I think smf just makes it harder to administer Solaris machines because it now does this thing that no one else does. After all, FreeBSD just got init script support in 5.3.

I don't want to live in a monolithic OS world. But, as a user and developer, I do want some constants so that I can remain sane as I hop from platform to platform. For example, a Unix system without zsh would be useless for me. I'm reduced to 'Newbie Level 1' when I see tcsh. But, give me zsh and I'm off doing my black magic. The same goes for init scripts: if I have to specifically think about how the system is initialized, I'll go nuts. That needs to be standardized as much as is practical!

Paradoxically, this very point is where I think OpenSolaris could play a role: help make innovation spread in a consistent fashion. How about developing a portable variant of smf under the auspices of OpenSolaris that works on Linux, FreeBSD, Mac OS X, and Solaris? A framework for starting processes doesn't have to be platform-specific - in fact, it's likely not at all. If smf appeared on every platform, then, as a developer, maybe I'd buy into its complexity because it'd work on every platform I need to support. That would justify my time learning those features. But, if it only works on Solaris, as a developer, I'm not inclined to learn that much more about the cool features and make my program support smf.

Of course, Sun appears to have written smf files for httpd - have they contributed them back? I'm also sure Graham would love help with packaging httpd on Solaris properly. More frequent interaction with the upstream open-source developers is something that would be valuable for the Solaris team to consider as they launch OpenSolaris.

[/software/solaris] permanent link

Game 9: A surprise: Chan Ho didn't keep the ball in the yard and won.

Figgins hitting a HR? Although, he did hit it fairly well. But, the #9 guy in the order with a shot like that? Yikes.

I think Byrd just got tired and was clinging on for dear life by the end.

Chan Ho pitched fairly well. So, far the starting pitching (except Young) has delivered beyond expectations. But, it's the bullpen that's been really rocky. Again, help is on the way there, we hope!

Gonzalez HR was a shot: this kid can hit. Buck had better put him in the lineup for tomorrow and the next day and the next...

I'd bet that Young's triple and Teixeira's double are homers if they were hit in July or August in TBIA. But, since it's only April, they didn't go the extra three or four feet to go over the wall.

Blalock's getting some walks which I think is a great sign. Patience. Very nice. Good sign of a maturing hitter.

Soriano made a wonderful defensive play in the top of the 9th. Yes, you may need to re-read that sentence again: it doesn't happen often. Heh.

Cordero gave up a 2-run single to Erstad. Has Erstad become Cordero's daddy?

[/sports/rangers/2005] permanent link