Timeline Chart or Graph of GNU/Linux Adoption? 49
DNAman asks: "I'm preparing a presentation for the use of GNU/Linux in the biological sciences. One recurring comment that comes up is 'Linux is not mainstream, why should we be interested in it?' While we could debate the definition of mainstream, I think it would be more productive to illustrate the trend in use / adoption of GNU/Linux as a platform. Do any of you have decent data sources for this type of trend?"
Unbelievable. (Score:5, Insightful)
Mainstream is for the unwashed masses (or maybe the washed masses, depending on the accuracy of the Linux geek stereotype). Mainstream is for people who want to surf the web and forward letters to everyone on their email list. If they never see how their computer works, so much the better; and mainstream continually works toward that goal.
Linux, on the other hand, is developed by people who have a similar mindset to scientists. It's a system that allows and encourages experimentation, and reinforces the truth that there is never only one way to do something. There are no artificial limits set to prevent you from getting uncomfortable because you're not sure what you're looking at.
It's also the closest thing to an objective OS you can get. If results are called into question, you're out of luck with Windows: "Oh, it must have been a glitch. Let me reboot this...." With Linux you can figure out what's going on behind the scenes.
I can't imagine scientists who willingly reject going a different direction from mainstream. Science is not mainstream.
Easy. (Score:5, Insightful)
That is, nobody will ever respect a theory untill it has undergone peer review. Other scientists scrutinize the theory, and try to disprove it. If they fail to disprove it, then it becomes accepted as the best theory.
In Linux, it's much the same. Somebody writes a program, and in doing so, claims that their program is the best way to solve that given problem. Other programmers will scrutinize the code, and try to find better ways of solving the problem. They can develop entirely new ways that replace the old ways, or they can incrementally improve the current ways. Either way, the end result is a higher quality of code, in a general sense.
I mean, Linux Just Makes Sense (TM) for the scientific community. They both heavily rely on peer review to ensure that things are reliable and trustworthy.
Would anybody trust a scientific theory that was developed by secretive scientists who won't publish what experiments they did to come up with their ideas? If we can't verify that what they did was the right way to do things, how can we trust their results? Of course you don't, which is the same reason that proprietary software is so crappy. You can't see their methods, only their results, and the results are often sub-par.
Numbers won't help you. (Score:5, Insightful)
Any such sources are inherently biased against linux.
The standard way for measuring the market share of a given operating system is to look at the vendor's sales. This has tons of problems.
First, linux ISOs can be freely downloaded without any sale taking place, so lots of people can (and do) install linux without showing up on any company's sales figures.
Second, even if you buy a boxed set of RedHat 9, you can install that on as many computers as you want. So if RedHat says they sold x copies of their OS, that could easily represent 2x or 3x installations of the OS.
Third, all new PCs come with Windows preloaded, and count towards Microsoft's market share. Even if the first thing you do with your new PC is install linux on it, it's still a Windows box, as far as Microsoft's sales figures will show.
So, sales figures are totally unreliable in gauging anybody's market share. You could turn to Netcraft, which is amazingly accurate... except that it only cares about servers. Desktops don't even show up at all.
If you want numbers, you're totally SOL. Even if you found the numbers, they don't mean anything. Ok, I'll tell you the numbers right now: Linux has 50% of the market. I pulled that number out of my ass, but it's about as reliable as you'll find anywhere else.
Re:Unbelievable. (Score:2, Insightful)
the code will give the same result no matter what operating system it is running under.
Unless the code is being timed. A scientific program has little use if an inefficient scheduler and an antivirus program eat all the CPU cycles.
Re:Youi need to ask yourself... (Score:2, Insightful)
Anyone who says something like this is just looking to argue (oops, I have been trolled). Computing platforms have been rapidly changing since the invention of the computer. I would peg full upgrade cycles at around five years on average (and interestingly this is how long it takes to fully depreciate the value of a computer purchase according to the IRS). This means that about every five years you are fully reinvesting in new hardware and software. Some software packages allow for (still expensive) upgrades, but that's small comfort against all the other expenses involved. And in a larger environment, you are likely to constantly be retiring old machines and replacing with new. So it's not likely you even have a single "platform" that "works". The problem here is how to explain a migration to Linux as a way to reduce future expenses, increase productivity, or whatever-- de facto migration to the next upgrade cycle of the current platform should require just as much justification.
Find out about usage at your institution! (Score:3, Insightful)
Re:Youi need to ask yourself... (Score:2, Insightful)
As for the rest of what you said, the types of systems you're talking about are large mainframes-- not something any sane person would be migrating to Linux any time soon. And I think you're wrong about payroll. PeopleSoft's most recent versions are hardly seven years old (although there may be seven-year-old code in their code base). In fact, given their whole company was founded in 1987 and is therefore only sixteen years old in the first place... I'd have to say your information is woefully out of date. And don't even get me started on the financial services where the ONLY systems that last more than five years are the big COBOL transaction systems that handle the basic account activities. But then, again, those are not likely systems for Linux migration. The thousands of desktops, clients, analysis databases, etc, are.
Re:Youi need to ask yourself... (Score:3, Insightful)