Slashdot Mirror
Linux Foundation Puts the Cost of Replacing Its Open Source Projects At $5 Billion
chicksdaddy writes: Everybody recognizes that open source software incredibly valuable, by providing a way to streamline the creation of new applications and services. But how valuable, exactly? The Linux Foundation has released a new research paper that tries to put a price tag on the value of the open source projects it comprises, and the price they've come up with is eye-popping: $5 billion. That's how much the Foundation believes it would cost for companies to have to rebuild or develop from scratch the software residing in its collaborative projects.
To arrive at that figure, the Foundation analyzed the code repositories of each one of its projects using the Constructive Cost Model (COCOMO) to estimate the total effort required to create these projects. With 115,013,302 total lines of source code, LF estimated the total amount of effort required to retrace the steps of collaborative development to be 41,192.25 person-years — or 1,356 developers 30 years to recreate the code base present in The Linux Foundation's current collaborative projects listed above.
To arrive at that figure, the Foundation analyzed the code repositories of each one of its projects using the Constructive Cost Model (COCOMO) to estimate the total effort required to create these projects. With 115,013,302 total lines of source code, LF estimated the total amount of effort required to retrace the steps of collaborative development to be 41,192.25 person-years — or 1,356 developers 30 years to recreate the code base present in The Linux Foundation's current collaborative projects listed above.
What is the cost of the QEMU code?
Is this like that facebook meme where if you added up everything a mom does in a day, it would cost you ONE BEEEELION DOLLARS a year?
They are presuming that all of their projects are equally valuable. The GCC compiler, for instance, is widely used, and it's disappearance would put a large hole in the software world. Gnu Hurd, on the other hand.... if it disappeared tomorrow, would anyone even notice?
If every corporation which relies on Linux as part of its infrastructure had to buy or build every piece of technology required to replace Linux, I should think on a global scale it would be far more than that.
Because a lot of that effort would be duplicated by multiple companies .. and of course the patent litigation by all of the players who seek to claim they invented some piece of technology which predates them.
I can believe $5 billion in this quite easily.
Of course, I can't read the paper since I need to fill out some fscking form from, and that's not happening.
Pity the Linux Foundation doesn't believe in open information.
Lost at C:>. Found at C.
For those who don't know, COCOMO is an algorithm that was developed in 1981 by Barry Boehm for estimating the cost of building software (typically in person-hours). The numbers in the article were generated by the basic COCOMO calculation in David Wheeler's free SLOCCount toolset.
One drawback is that SLOCCount uses the basic COCOMO calculation, which is based on historical data gathered by Boehm in 1981. Here's a COCOMO-81 calculator in case you want to play with your own code. Sometimes its estimates are pretty good, but I've sometimes found that applying line counts from my projects in some modern languages (especially functional ones like Scala) throw it off. That could definitely affect the "1,356 developers 30 years" estimate in the article.
Wheeler has a good discussion of COCOMO in SLOCCount if you want to learn more about it.
Building Better Software
"But my time is worthless, that's why I use open source."
Seriously, though... the world works on open source these days. I would say is another bogus calculation and the real harm would be incalculable.
Or take it another way... this is theoretical enough to be useless. Because the source is OPEN it's impossible to eradicate. You nuke one code repository and five more spring up in its place.
As always, since the very nature of these projects mean you don't have a marketing teaming going "rah rah" all the time in background like Microsoft does people don't KNOW the world runs on open source projects...
As much as I don't like "He who should not be mentioned," at least it is a contribution. There is some implicit agreement that the action is worth something or else it would never have been done.
lotta rogue apps come running thru here. Be a shame if one of them, you know... No, no - I'm not saying nothing, I'm just saying. (It's a joke. What they've accomplished in software is stunningly good.)
"Win treats sysadmins better than users. Mac treats users better than sysadmins. Linux treats everyone like sysadmins."
Umm, everybody uses 1 billion = 1000 million these days, the 1 billion = 1 million million definition is long obsolete.
So you're admitting America is part of the civilized world?
That's a rare concession coming from the poor inbred eurotrash element.
"Some bits are more equal than other bits." - George Orwell
Oh great, now the information on taxing FOSS has been created. The government can now demand to know how much FOSS a person or company is using and then tax them on the "value" of it even though you paid nothing (and for individuals not earning anything from it).
If you don't know how a government does imputed income, let me cite an example that almost got done. The current US "regime" wanted to charge home owners who had taken a mortgage on a house years ago and were making relatively small payments by current rates, the difference between what their house would rent for (if it was more than the mortgage payment) and the mortgage payment as imputed income. Yes, if you were paying $500 a month and the house could rent for $1500, you would have to add $12000 to your annual income in "imputed" income.
Now take that and apply it to FOSS. One single distribution in use could be imputed to be worth thousands of dollars. Microsoft, Oracle and others would love to latch onto that and use it to kill FOSS if they can only get the government to do their bidding. The ability to do this in other countries may vary but probably not by much.
They said it would take approximately 30 years for approximately 1300 developers to get there. We know because we have an idea of how things evolved that estimate is actually a bit short. Some of that codebase is about 30 years old, and well more than that many developers have contributed. Things have been done, discarded, redone. The estimate is actually a pretty optimistic one that assumes the developers get it 'mostly' right the first time when actual history has had many many dead ends that caused a total rethink. One would expect the same out of a private endeavor. So there's some balancing out.
XML is like violence. If it doesn't solve the problem, use more.
Well, sure. America is part of the civilized world... except for the rampant genital mutilation.
> the price they've come up with is eye-popping: $5 billion.
Apple claims to have sold 13 million iPhones on launch weekend. Assuming an average price of $800, that's 10 billion in two days.
So this doesn't impress me much. Or at all. I suspect its at least an order of magnitude higher than their estimate.
Since most companies would either develop proprietary solutions or buy at a substantial markup from an establish publisher, the actual cost to replace all that software would be much, much higher. And if an established organization offers a replacement, it will likely have competitors---which again gives a duplication of effort, even if it is much smaller duplication than proprietary redevelopment of the functionality. This is not addressed at all in the paper.
They do acknowledge that failed or superseded code is not included in their analysis, and there was certainly developer time spent on code that is not a part of the project, either because it was culled or never made the cut to begin with.
Given both of those factors, the $5 billion figure is a very low best-case value. The practical cost of replacement would be monumentally higher once the mundane practicalities come into play.
---
According to the latest ruleset, this post should be modded as Vorpal Flamebait +5.
But I guess if enough others save hundreds as well that figure can easily be in the billions globally. Still, it is the hundreds saved that are important to me.
Time is what keeps everything from happening all at once.
Just simply calculating 5B / (1356*30) you get approximately $122K per year. Bull shit. No fucking way 14 year old Billy from Boise in his mom's basement is making any money, let alone 122 grand.
Or, more likely, switch to FreeBSD and forget Linux ever existed.
Yes, who indeed? GKH? Alan Cox? Lennart P? Kay Sievers? rms? Linus?
In France we use the long scale but the term "billion" is almost never used. The short scale billion is called milliard and for the short scale trillions, we simply use thousands of milliards.
Short scale trillions are so big anyways that they only seem to be used to express national debts. In science and computing, SI prefixes are preferred.
Trying to cheat me out of moneybits.
1 billion = 1024 million.
The costs also need to consider all the dead ends and bad decisions. How much code was built that was never injected into the system. I suspect that many Linux developers have conjured up some really long and interesting code that they then never submitted. They threw it out and started again.
Then there is the quality of the programmers. The few programmers that I have met who contributed to the Linux codebase were pretty damn kickass. Thus hiring them would not only be expensive but really hard. Most of them wouldn't work for most companies as they know they are the elite of the elite and can pick and choose their surroundings.
Also Linux contributions are often a resume builder. Thus many junior but very very good programmers will do some Linux contributions which then makes them look cool. The reality is that they don't want to work on Linux but want some other job, such as the games industry. This is a double problem. Some company hiring for their 5 billion dollar project would never have hired them because they had crap resumes, and these kids didn't want to work on Linux and thus wouldn't accept a job working for a big boring company building an OS.
Then there is the urgency factor. Many critical tiny bits of Linux were built by people with a specific problem. They didn't have a Linux driver for their 10,000 machines with the L257B Arcnet card. Thus they dove in and modified the driver for the L256A arcnet card just enough to make it work. But where would that kind of bug/feature have been prioritized by a corporation? Plus again the person doing this brought a skillset that was obviously very good for that problem but otherwise might have been a terrible hire for the project.
Then there are the various OS distributions that compete. Not all Linux decisions have been good ones. Thus different distributions follow different paths resulting in winners and losers for various aspects of the system. Over time the winners end up spreading across the distributions and the losers just sort of fade away. Even the classic Gnome vs KDE has resulted in each becoming better. So one must count the costs of developing both Gnome and KDE.
This last one even extends out to other Open Source OS projects such as BSD in that code from that project end up in Linux as well as providing competition.
Then there is the whole build the wrong thing problem. Linux has evolved steadily to meet the demand of its users. But a corporation would build a product that would meet the demands of its marketing department. Thus any corporation building Linux wouldn't build Linux. They would typically build something like Windows or OS/2; Operating systems that were designed to create an ecosystem for selling other crap made by that company and locking their customers in.
So while it is interesting to say such a huge number, I personally think that the number would be far far larger as to put together such a talent pool would probably be a mega project in itself over and above the actual paying of that pool and the other development costs.
So you could take about 1/3 of NASA's yearly budget and recreate it all from scratch. Just think of what could be available if NASA was defunded for 5 years and that money went into a national open source development project where everything created would be free to the world.
Of course, imagine if we did the same with 1/2 of the US military budget. I suspect you'd run out of developers to hire before you ran out of money. And the pay for developers would be higher than airline pilots.
Ninjas don't carry tic tacs
Amount of testing needed to achieve the same level of reliability and interoperability ? Could easily top 500 billion dollars.
No, it is not hyperbole. Look at the expense most companies are going through to maintain ageing old code running in mainframes or the code running on WinXP and IE6 and ActiveX control. If you look at the man years used to develop that code, it might be X. To rip it out and replace it? It has no relationship to X. It all depends on how many other modules depend on it.
It is true for everything, not just code. Cost of constructing Tappan Zee bridge would bear no relationship to the cost of blowing it down and rebuilding a replacement in situ. The cost of service interruption alone is going to exceed the original investment by several times over.
sed -e 's/Chuck Norris/Rajnikant/g' joke > fact
I never found COCOMO particularly useful for cost predictions because you guess SLOC Instead of guessing FTEs. Reversing the process when you know the SLOC still has problems because COCOMO is nowhere near being reliable if the SW involved more than screen to database and reverse, such as significant math.
a nice upgrade to LLVM.
And all of the architectures that are not supported by LLVM are then screwed...
That's part of the upgrade.
It's only 115M lines of code.
My calculation on that comes out closer to $1.3B for a 5 year project to replace all of it.
With a much smaller number of highly dedicated people who are 3X as expensive as the average software engineer in Silicon Valley, I think it'd be possible to drive the number down closer to $790M and 2 years.
The people would need to be dedicated, and the project would need to be driven by (in effect) a dictatorial ass whom everyone has agreed to follow to the ends of the Earth. In other words, a rather strict hierarchy.
Oh yeah... the project would have to be operated Cathedral style, and not use most "agile" techniques.
I must confess... I'm kind of partial to my dirty bits.
"So long and thanks for all the fish."
All you need to do is hire 494,304 programmers (41,192 x 12) and you could replace it all in a single month. Right??? See how simple math works.
For the first time ever I wish I had mod points!
Equal Rights, Representation, Education & Welfare
What would it cost for the US or UK govt to try to duplicate this code base?
I only look human.
My mother is a halfling and my dad is an ogre, so that makes me an Ogreling
Just because it would cost $5 billion to replicate the c code doesn't mean its value is $5 billion. Some could be worth a lot more that the replication costs, and some total worthless despite huge replication costs.
I'm a consultant - I convert gibberish into cash-flow.
I am a desktop end-user of LInux, without any anti-systemd feelings because for me there really hasn't been much difference. But I think the above post is hilarious and gets the references right.
You do need something like this too: "Lennkor had the aid of Fedoriant. in the poisoning of the Process Trees that enabled him to take the Sysvmarils.
We all understand the mythical man-month issues here, but Microsoft alone employs around 50,000 software engineers (Google 30,000). If everything done at the linux foundation is equal to less that the output of Microsoft for one year (or google for less than 16 months), there may either be a problem with your numbers, or with the entire model of open source. I'm going to assume the issue is the models...
That works out to 8238450 work days of programming (presuming 50 week years.)
That means they only expect a programmer to produce 1396 lines of code per day.
It would seem they're over-estimating the cost of their projects -- even back in the early '90s an "average" programmer produced 2000 lines of code per day, and that was before the advent of most of the modern debugging and IDE technology that speeds up the process, included time for builds which used to run for hours or days instead of minutes, and involved a lot of manual processing to integrate code from different developers because versioning tools were so primitive (RCS days.)
I do not fail; I succeed at finding out what does not work.
So, in a commercial environment, it would cost $5 Billion of commercial developers being paid a proper wage. I get that. But in this day and age, if you wanted to build that sort of thing, you wouldn't hire developers commercially. You'd create an open source project and let the developer community at large assist in your project.
In doing so, it would cost far less. I cite, as my proof-of-concept example, an organization called the linux foundation, which has 115,013,302 total lines of source code and didn't spend $5 billion to get there.
I used to wonder how I'd dig up my shovel, without my shovel. Now I wonder how I buried my shovel, without my shovel.
That seems low honestly. That's pocket change to the big fish.
Chick Norris can rewrite the entire kernel in a day . Without a single bug.
Hans Reiser?
(1.21 gigawatts) / (88 miles per hour) = 30 757 874 newtons
Is she Chuck's sister or his drag act ?
Unicode killed the ASCII-art *
Interesting. In Afrikaans Miljard (which is pronounced almost exactly like milliard) is only a hundred million.
Unicode killed the ASCII-art *
In France we use the long scale but the term "billion" is almost never used. The short scale billion is called milliard and for the short scale trillions, we simply use thousands of milliards.
Ditto for Italian. Long scale terms seem to imply a billiard is 10^15. Carambola!