Qantas ditches Linux for AIX

> Qantas had no problem putting its systems on Linux.

Point taken. I was arguing the more general case in my response to number6x however, not something specific to Qantas.

>I was arguing the more general case in my response to number6x however, not something specific to Qantas.

Sure. I think, however, that Linux really has grown beyond the guinea pig stage.

More likely a matter of businesses being VERY conservative with their bread and butter systems. Same factor will apply to companies who won't move from mainframes to, say, Solaris -- the safest thing to do is nothing. When you can't do nothing, the next-safest thing to do is whatever most resembles what you're doing now.

When you've got billions riding on your systems, you handle them with care.

> More likely a matter of businesses being VERY conservative with their bread and butter systems.

Same thing from two different viewpoints, Dino. You say tomato, I say...

Quoting: dino: When you've got billions riding on your systems, you handle them with care.

My point exactly. So why on earth would the manager of such a system ever decide to "download the latest kernel" and compile it himself? Answer. No way he would ever do anything like that. Mr manager runs the official, vendor supplied kernels, thank you very much!

So I'm still having a hard time seeing any concrete way in which the admittedly rapid pace of kernel development could have any bearing whatsoever on already deployed systems, such as the 30 SLES servers which run our infrastructure here.

>So why on earth would the manager of such a system ever decide to "download the latest kernel" and compile it himself?

Ummm....

1. Using a distribution's official kernel is no guarantee against the kind of scenarios I have describe.

2. Your experience in the great wide world seems to be pretty limited. There are shops who do things because they have little choice in the matter.

Quoting: dino: 1. Using a distribution's official kernel is no guarantee against the kind of scenarios I have describe.

The kind of scenario you describe would never arise in a serious production environment. Think about it. You've got the company's business riding on your transaction server, you're telling me that one fine day the manager will decide that he wants better device support for some newfangled gadget, so he'll roll the dice with a new self compiled kernel? No, I don't think so.

Quoting: dino: 2. Your experience in the great wide world seems to be pretty limited. There are shops who do things because they have little choice in the matter.

Thanks for the vote of confidence - perhaps I can't boast as many years IT experience as some of the dinosaurs here, but I've deployed linux (and other unix) solutions in a number of businesses, from one-man shops selling discount software online, to multinational corporations with data centers spread out over 400 miles. I hold down a day job as a unix system administrator with a fortune 100 company, and I consult on the side, for small and medium businesses.

I have to say, this is the first mention I've ever heard about my lack of experience.

Something to consider: I read a book recently that said "There are often two reasons to do (or not do) something. One that sounds good, and the real reason."

Perhaps they got a new IT boss who rather then look un-knowledgable about Linux or perhaps being too lazy to learn the Linux system (the real reason) pushes to replace it with something he is more familiar with because it is more "stable" (sounds good).

This happened in my company after I sold it, except in that case they hired a Windows guy to run the system (the first stupid decision, why would you hire a windows guy to run a non-windows system?). Months later I stopped in to visit and he had converted the whole business to his familiar Windows rather then have to look stupid (appearently it's ok to be stupid, just not look stupid) or learn anything. He can now justify his existence by dealing with all the Windows problems he is comfortable with that I never had to deal with. (I have to admit that after I switched over to Linux, I came to realize that most of the issues I had with "computers" were really issues with Windows. The problems just disappeared)

Now whatever their situation is, I would be willing to bet that the stability thing is just something that sounds good.

Just a thought.

>I have to say, this is the first mention I've ever heard about my lack of experience.

Well, glad to be the first to make you aware that the world is a large and varied place. It sounds like you fall victim to a common conceit -- so common that it inflicts us all: if you ain't seen it, it don't exist.

It's not a knock on you. It's unavoidable.

It's been 20 years since I worked at EDS, and it still frames the way I look at IT shops.

> The kind of scenario you describe would never arise in a serious production environment.

If they're already running a custom kernel, and the hardware changes, they'll probably have to compile a new custom kernel. Or are you arguing that no one would ever run a custom kernel in a "serious production environment"? That argument doesn't sound plausible to me.

>That argument doesn't sound plausible to me.

It's not.

I'd forgotten about it before, but I remember reading an article about some sort of sorting machine used by the post office (and, I believe, only by the post office) that required custom drivers. Though not the norm, there are many companies/organizations throughout the IT universe that have "outside of the norm" needs. Heck, Google! They modify the OS pretty heavily. Of course, their mods are aimed at surviving unreliability, but, hey!, it's an example of custom kernels in a production environment -- and, yes, a web site/search engine IS a production environment if it is your company's product.

Quoting: jdixon: If they're already running a custom kernel, and the hardware changes, they'll probably have to compile a new custom kernel. Or are you arguing that no one would ever run a custom kernel in a "serious production environment"? That argument doesn't sound plausible to me.

Nice try, but how can you say on the one hand, that these serious production environments are ultra conservative, and schedule changes months in advance, and then turn around and say that they fly by the seat of their pants and run experimental kernels? You can't have it both ways.

If they are conservative and uptight, they can be just as conservative and uptight with SLES running on HP/Compaq hardware as they can on a hpux superdome or a sunfire e12k. Such a linux server will stay up for 1500 days, or until you take it down. Your call.

OTOH, if they are technically skilled and confident, and have no problem tweaking the kernel and writing custom code, (google, slashdot, amazon) linux is an even better fit than it's proprietary cousins, since it is ideal for such scenarios.

So, no matter how you cut it, linux works. I say the Qantas thing was a PR stunt and the result of some sort of agenda, not a genuine technical issue.

> and then turn around and say that they fly by the seat of their pants and run experimental kernels?

Where did I say they might ran experimental kernels? I said they might run "custom" kernels. That's not even remotely the same thing. A custom kernel is one compiled to your specifications, nothing more or less.

> OTOH, if they are technically skilled and confident, and have no problem tweaking the kernel and writing custom code, (google, slashdot, amazon) linux is an even better fit than it's proprietary cousins, since it is ideal for such scenarios.

Well, at least we agree on something.

> So, no matter how you cut it, linux works. I say the Qantas thing was a PR stunt and the result of some sort of agenda, not a genuine technical issue.

That's always possible, but there's simply no way we can know for certain. It's always possible that Qantas hit on some obscure kernel bug and weren't able to get it resolved to their satisfaction. It's also possible they were lying through their teeth. Or anything in between. We have absolutely no way of knowing.

Quoting: jdixon: That's always possible, but there's simply no way we can know for certain. It's always possible that Qantas hit on some obscure kernel bug and weren't able to get it resolved to their satisfaction. It's also possible they were lying through their teeth. Or anything in between. We have absolutely no way of knowing.

Call it an intuitive gut feeling backed by years of experience.

> Call it an intuitive gut feeling backed by years of experience.

OK. But stating that it's an opinion doesn't really back up your argument. :)

My intuition agrees that there's more to the story than was covered in the article. We're just disagreeing about the details.

Quoting: jdixon: OK. But stating that it's an opinion doesn't really back up your argument. :)

Not an opinion - a flash of intuition. Big difference. I won't even try to explain that, but it works for me. end of story.

OTOH, the arguments I made were an effort to debunk some of the straw men I saw bandied about here (e.g. that the rapid pace of kernel development somehow affects the stability of already deployed linux systems)

> the rapid pace of kernel development somehow affects the stability of already deployed linux systems

Who made that point? Sure wasn't me.

Quoting: dino: Who made that point? Sure wasn't me.

Wonderful! we're on the same page then.

>Wonderful! we're on the same page then.

Page ain't the problem. It's the chapter I'm not sure about...

Quoting: dino: Page ain't the problem. It's the chapter I'm not sure about...

Thank you. It takes a big man to admit when he's wrong.

>Thank you. It takes a big man to admit when he's wrong.

Sure does. When I am, I will.