Do you think that MATLAB is expensive?
Paulo Silva
on 14 Aug 2011
Latest activity Reply by Christopher Hartline
on 12 Aug 2022
We all know that MATLAB is probably the best software for engineering purposes, I think it's a little expensive unless you have it for free on your school or place you work, please share your opinion about MATLAB cost, including toolboxes, student versions... is it that expensive?
82 Comments
Time Descending
The only way that Student and Home licenses will not be used to earn income would be to cripple them to the point that they could not be used to earn income, at which point they would be nearly worthless.
The fact that Mathworks has continued to make money from Matlab for many decades speaks to the value of its current business-model as well as its management over time. That doesn't mean that the Matlab market operates the way that Mathworks wants it to. But it is working.
The only question is whether or not that will continue in the face of the OS movement.
And tht is a question for every company that develops software with restrictive license terms, even for OS re the GPL.
So here's the thing. Bugatti develops, makes and sells vehicles which are based on high prices and a market of owners who are happy to buy their products. Bugatti is the "boutique" label for Volkswagen. It's not like there is not a market for "expensive" software products. Even Microsoft knows the wisdom of giving-away products at the low end to generate revenue at the high end. The question is where does Matlab sit in the Mathworks software lineup and why even entertain the interests of users (who may or may not be legitimate) to make Matlab more compatible with OS tools. That's like the directors at Bugatti saying "hm, maybe we should sell a $5k version of the Chiron, or at the very least find some maker of cheap aftermarket parts for expensive cars and collaborate with them". This is why I love the current Matlab model of toolkits and add-ons for an additional price. Customers pay to leverage their investment in the basic Matlab product, those toolkits have a real-world market value that should fluctuate together with their price. Those toolkits are worthless, supposedly, without Matlab. That's exactly how it should be, in my opinion. I don't see the wisdom in trying to control or even influence the Matlab market through the licensing of product versions that are technically the same, that's like asking the market to violate those licenses. But by the same token so is adding value to the product that is commensurate with the price. This is a problem with the entire software market: it is crucial for developers to be able to control who runs their software but by the same token they need information to decide who should be able to run their software. That information might be as valuable as the software itself. Of course a market will develop to disable that control, but that's par for the course. At some level it's also a battle between Matlabs license security and integrity and hackers who want to free Matlab code to be run at will.
Using a Student license to earn income is absolutely against the MATLAB license terms. Same for a Home license or MOOC license.
Formally speaking, it is against the license terms for Academic licenses as well. I am not sure if there are exemptions for cases where internal university regulations require each group to act as a "cost center" that charges other groups for services. If it would normally be entirely within the license terms to write some software for internal use, but internal regulations require that the group writing the software bill the receivers, then I do not know how Mathworks would handle that.
Doesn't matter
You either need it or you don't need it
If you don't need it, then why are you worried about how much it costs
If you do need it, then you can factor the price of a license into the charge to customer
So your only problem is if you do need it but you can't spread the cost out among your clients.
And if that is the case then take a class at a local college that offers free Matlab to its students (Matlab will give you a student license if you proved them with a student email) & do your work there.
Or learn to accomplish your tasks without Matlab.
If you are not earning money from your use of MATLAB then you may be eligible for the MATLAB Home license. https://www.mathworks.com/products/matlab-home.html
It would be nice if Matlab had a low-income as well as a student price. I'm 72 on a low income (EBT card) but don't think my brain has decayed yet and want to do some work in Graphs and AI. But the regular price is way too high for me. Matlab should let you provide proof of income, such as an EBT card number, for the student price. A lot of us are self-learners.
Expensive but very good software support.
Yes, it is expensive. But the quality of support you will get from this software will justify the cost.
MATLAB is probably expensive for a small company. For a large company, I think it is relatively inexpensive. If I ever need to contact MATLAB support to figure out a function call (e.g. once I had to interface with C# before), usually a reply with example code would be in my email the next morning. Compare that to having say Python and Google searching, StackExchange questions, etc. and waiting for a reply. So if you value productivity the benefits are great. I also am a regular Python user and despite liking the language, MATLAB is far easier to get work done in. The syntax is cleaner than other languages for mathematical work. The data structures tend to be easier to work with and make more sense. Compare to Pandas, NumPy - sometimes I play around for 10m trying to figure out proper indexing for a complicated call. In MATLAB usually the help menu has an answer. It's just a good commercial product.
Company size could matter, if you're a startup company you may be eligible for the MATLAB and Simulink for Startups program.
"MATLAB is probably expensive for a small company. For a large company, I think it is relatively inexpensive."
The corporate size doesn't matter at all, what matters is the budget for items like Matlab and the pros & cons of using Matlab for development.
Matlab has been around a long, long time in a market where hundreds of other tech companies have come and gone. My guess is that they know what they are doing. The question is not whether Matlab is "expensive" or not. The question is whether you need to buy it or not. I wouldn't spend $50 on a Matlab license if I didn't need it.
I have used MATLAB and MAPLE. I think MAPLE is cheaper than MATLAB when purchasing more than one license and the base MAPLE include 70% of the toolbox that MATLAB would have required.
Vaishali Scientific Centre:
I deleted your quotation because price quotations are often considered confidential information.
Your quote was for 10 concurrent academic licenses of a series of toolboxes. The pricing you were given appears to be about 40% or more lower than US or Canada academic licenses, I calculate. The pricing you were given appears to be less than 12% of the price I would have to pay for 10 licenses for those toolboxes, as I am not Academic.
Concurrent licenses from Mathworks are the most expensive form of licenses: in your case they would allow any 10 people from your institution to use the software simultaneously. If you are looking for having 10 specific people use the software, then you could ask about Networked Named User licenses, which are less expensive than Concurrent licenses.
Depending on what your people need to do with the software, you might find that although eventually each user might use a particular toolbox, that it might be uncommon that more than (say) 4 users at a time are using one of the toolboxes, where-as a different toolbox might often be in use by (say) 7 of your people and it might turn out all 10 need that toolbox a fraction of the time. In such a case you might consider ordering only (say) 5 or 6 of the one that is not typically going to be used by most people. The Concurrent licenses and the Group (Networked Named User) licenses both keep "pools" of licenses, and will allow people to check out licenses as long as not more than the configured number of people are trying to use that toolbox simultaneously.
The disadvantage of shared licenses like this is that the licenses are considered to be "borrowed" from the time they are first used until the time the MATLAB session ends, so especially if people do not log off when they go home, you can end up chasing people to quit their MATLAB session in order to free up the toolbox license of something they used a while ago but are not actively using now.
MATLA Perpetual license is so much expensive. Sales executive is not coperative.
I tried to purchase Simulink, but was told I needed to become current on my support service tithing first. So it adds another $2K to the price. No Thanks. Time to go find another tool!
Hobby user here... If you think it is expensive, try out an open-source version... GNU Octave . It's free. It's very similar, albeit not as supported, and you get the source code to compile and use as your heart desires. I was using Octave for a while-- and it's pretty good. However the business case for Matlab got me to finally fork over the $$$ even though it does the same basic functionality. Here's how: I was intent on developing my own home-grown toolbox(es) to do a specific task (neural network functions with specific interfaces to play around with...) It would have taken me a month to do this on my own on Octave given my schedule. The home grown may have been tailor-made and "better" in some respects.. However, for the $150 home license (glad they finally added this option!), and the $50 for the NN toolbox.. (and a few more fifties for some other toolboxes I found helpful) I can't justify NOT spending $200+ to save many weeks work. The guys at Matlab are awesome. They do a great job and the support is fantastic. If you are hell-bent on saving every nickel-- use Octave. If you can rub a couple hundred bucks together for a home license, I'd do that (and I did). --And if I ever had a business case where I required it for business, I'd pay for the full license if it were using it in that capacity. There are lots of options.
...still waiting on you guys to sort out the issues with the IP stack on Win10 machines (a growing issue with security concerns) so I don't have to use a pcap stack to do network i/o at the packet level.
Not that this is a major complaint, I assume that the Communications Toolbox will work with the pcap stack if I try to use it that way instead of writing an interface between it and Matlab. Or SSH if necessary so I don't have to write something like scp.
In a place I used to work, I had a co-worker who was an ardent FOSS person. They apparently believed it was better to devote a team of three PhDs and two programmers for two years to create an open source version of something rather than spend $50 per year on a software license. It boggled my mind.
That's very expensive when it is not your main engineering tool but convenient environment for post- and pre- processing and "quick and dirty" proof-of-concept models. I might get away with the basic licence only (which has a fairly affordable price of £375) but I need to make sure that modules covered by basic licence contain all thh functionality I need
Anastasia Makarevich:
The purchase of the Home license includes one year of Software Support Service, which permits you to open support cases and to download and use any new versions of MATLAB that come out during the term you are covered. You will be permitted to renew the support service, at a price. The license is "perpetual", meaning that it will not expire and that you will be able to continue to use the version you paid for, but you will not be entitled to new versions after your support service expires unless you renew the support service.
You should contact Mathworks directly with these questions. People here are mostly users and whilst they can give advice, this is a subject that you should want answers directly from Mathworks on. Someone here may give you well meaning advice that turns out to be incorrect.
I still don't understand the payment scheme. I've just bought a student license. Recently I received a letter from MathWorks that I would be able to download newer versions during the next year and the they will remind me when it's time to renew the license. I was sure I wasn't buying a subscription. Will they take my Matlab from next year if I can't pay (it's hard to predict the future)?
As far as I am aware Matlab Home has access to all functions in the base Matlab and if you add on toolboxes you will get access to those too with a couple of exceptions I think (e.g. I don't think Matlab Coder is available with a home license).
As a "retired" person I am happy that mathwork finally decided to have an Home edition. But it is sad that the Home edition does not offer the possibility to upgrade! Spending the money for the Home Edition and adding Toolboxes as the financial ability makes it possible is an effort well worth and justifiable at the reduced cost within the family budget. But the development related to technologies related to the design and modeling process linked to external proprietary hardware is taking place at a very fast speed makes the fact of not being able to update the tools painful.
Thanks! This looks like something I could recommend to readers of my book. I went to the Matlab Home site but it does not tell me what built-in functions are available. Where can I go to find out before I recommend it?
Thanks,
Dave
@David, there is the MATLAB Home Edition available for non-commercial use:
I first bought Matlab in 1989. It was on one 5 1/4 inch floppy disc. I was astounded at what I could now do. As time went on I kept updating Matlab until 2001 when I retired. A few years later I became a Matlab author and now keep reasonably up to date for free (I am now a two-time Matlab author).
I am frustrated that individuals who might buy my books can not afford to use Matlab. I have tried Octave and, although it can do elementary things, it stumbles badly when trying to deal with multi-second wav files and 3D graphics (which I want readers to use). A basic 64-bit version that contained the tools in the 1989 Matlab would be wonderful if it was priced reasonably, say, on the order of $100. Then I could recommend it to individuals not working for a large technical company.
Unfortunately, I do not think Mathworks could gainfully sell a small package like this and make money.
We also use MATLAB in my old company. We use it along with other expensive programs like CATIA and ANSYS. But we were able to cut back on the license cost by using software asset management tools. My former colleague uses OpenIT ( www.openit.com )for monitoring license usage.He's an IT guy there, he's mainly using it for reporting and facilitating IT chargeback. They say the software also has license harvesting feature. I'm not sure if they have a free program, but i think they have free demo. Anyways the good thing about it is that, it works even without the license manager. Haven’t tried it yet, but i maybe evaluating the Open iT software soon. Will let you know if it works for me.
Whichever way you look at it, MATLAB is expensive. But that's not to say it isn't good value. The value of MATLAB of course depends entirely on its use/user, and many users including myself would argue that MATLAB does offer very good value. As a total package, it is the best software available.
MATLAB is not the only expensive software out there, look at AUTODESK or ANSYS for example. Unlike AUTOCAD which has MICROSTATION, MATLAB is essentially free from direct competition. MATLAB's main competitors in the control sectors are National Instruments and Esterel Technologies, the latter of which just got bought out by ANSYS. Of these, in my oppinion MATLAB offers the best and most complete product and should therefore demand the highest price.
But I do wish there were more flexible licensing options available. For example, sometimes you may only require MATLAB for a short duration, for a fixed specified project.
I feel for those who want to buy additional toolboxes for old versions. This should be allowed, because some people may want to standardise on a specific release. Especially if clients or colleagues are working from the same version.
I also think it should be easier to patch MATLAB. For example, you can go to the bug reports page and download patches for many fixed issues, but this is an entirely manual, time consuming and sometimes complex process. IMHO there should be an automatic software update mechanism.
Lastly, I agree that there are too many toolboxes that are too much alike. I think Mathworks should continue to merge very simuliar toolboxes to simplify the product portfolio. Take the Signal Processing Toolbox and DSP System Toolbox for example... Or the MATLAB and Simulink Report Generators... Personally I don't think that's justified and only serves to complicate their portfolio.
I am glad that MATLAB is a commercial product, developers need to get paid and for those that rely on quality, there is simply no comparison to the free 'alternatives'. Just my 2 cents...
Paul
I also think it should be easier to patch MATLAB. For example, you can go to the bug reports page and download patches for many fixed issues, but this is an entirely manual, time consuming and sometimes complex process. IMHO there should be an automatic software update mechanism.
FYI, since this answer was posted MathWorks has begun offering Update releases that include fixes for a collection of bugs reported against each release. As an example, as of the time I write this there have been 7 Update releases for general release R2020b that have included hundreds of bugs fixes.
"Whichever way you look at it, MATLAB is expensive."
...well, there's absolutely expensive, and there's relatively expensive.
I think that's enough said in response to your comment.
Personnally no, I think its fairly priced. Mathworks developers need to be paid after all, and a reduction in fees will have an impact on the quality of the product - its not a listed company and there are no shareholders to pay dividends to. In 2010 The Mathworks Ltd had Sales of £22.69m and after costs had a Net Income of £290k only.
One of the advantages of matlab is its easy interfacing ability, so rather than say buy the datafeed toolbox, you can write your own Bloomberg API C# code and use it via matlab dotNet Interface, same with database toolbox - its pretty easy and quick in C#. Only buy a toolbox if really necessary!
R is extremely unreliable as the code coming from the user community is very prone to errors (without being rigourously checked for bugs) and not reliably supported (you may get a reply from author in 1day or quite possibly never), its a black box. And if you have to take the time to run through it yourself to check everything, then you may as well be writting it yourself. C++ of course is faster in execution but takes many many more man-hours to develop anything - this time is curcial for us.
These days people seem to more and more expect everything to be for free, I guess music and movie downloads have something to do with it.
@Mark Whirly
Reproducibility seems often to be a bigger issue in financial modelling. I've been in "Meetups" where none of the participants were too bothered about the precision, but were concerned to reproduce their results exactly in the event of a law suit. The same rule applies in other fields and is the reason MATLAB Version 2.01 can still be found running of a steam-powered 80286 PC in the corner of some offices.
For my own work, IEEE double precision is typically more than adequate, but it can be too easy to dismiss innacuracies as due to rounding errors when, in fact, there is a bug that should be addressed.
Hi Malcolm For my own application, I'm not too concerned with machine/truncation error (since financial modelling involves estimation, overfitting and misspecification aspects which far far outweigh this in my view). I agree Matlab does make an accuracy/speed tradeoff but some kind of tradeoff is an inevitability in all languages I think (bugfreeness & usability being the other dimensions of this tradeoff). The difference between Matlab and open R scripts is that I have the confidence of knowing that code has been checked multiple times by a team, and that they have full accountability for this - so I'm confident it has the best chance of being bug-free (and finding a tradeoff that adheres to a uniform approach rather than R which is big unknown). I recently attended a presentation by Jos Martin in mathworks where he describes their philosophy to development (maybe you'll find it interesting?).
@Jan
Thanks for this. Your FEX submission illustrates the advantage of open-source. Code available, documented and referenced. Therefore open to discussion/suggestion. Here's one suggestion: as commented in your code, there is a typo in the 2004 Ogita paper. It might be helpful flagging that on the FEX description - it would have saved me some head scratching.
PS A handy test of various algorithms is discussed at http://developers.sun.com/solaris/articles/fp_errors.html
Kahan's error compensation method is very nice. But the method to accumulate ina 128bit double described by Knuth is faster and remarkably more accurate, see http://www.mathworks.com/matlabcentral/fileexchange/26800-xsum.
"...if you have to take the time to run through it yourself..."
That is true for MATLAB too. It sometimes trades accuracy for speed, and when/how are not always documented. It also relies on third party products: "We rely on code generated by various C compilers and settings in various libraries, especially Intel's Math Kernel Library (MKL), that we do not control."( <http://www.mathworks.co.uk/support/solutions/en/data/1-79FEJH/index.html?solution=1-79FEJH>)
Take a simple sum:
>> x=rand(1,1000);
>> sum(x)==sum(fliplr(x))
ans =
0
So Kahan compensation is not being used. Is any?
In Octave you can select the 'extra' option. The docs tell us "For double precision inputs, 'extra' indicates that a more accurate algorithm than straightforward summation is to be used". It might of course be even more helpful if they also told us which.
While the argument of the netto income is reasonable, it is not exhaustive: TMW would sell more licenses, if Matlab is cheaper. A lot of users can aford a single release only, and because bugs are usually fixed by upgrading only, they obtain a limited stability only. So you get 2 scenarios:
- Matlab + 1 year service = 1000$ (fictive, absolute value does not matter here). With a budget of 2000$ for 2 years, a lab buys two licenses for 1 year, and lives with the old release including its bugs for the 2nd year.
- Matlab + 1 year service = 500$, lab has the same budget (of course): 2 licenses for 2 years are bought, TMW earns the same money, but the lab has a newer and more stable system in the 2nd year. And the free-lancer XY can afford such a license also => 500$ plus for TMW.
Therefore I do agree only (but at least) partially, that the netto income is a valid argument for the costs of the product.
Yes, I think it is expensive.
In New Zealand, it's about $2000 per toolbox. The compiler is over $8000. The work I was doing was going to rack up $20k in licensing. I had to make some sacrifices and squeeze that down to about $10k. And this is for a single user only.
If I was to do this all again, I would learn R or Python or just do it in C++. But being a developer, that's probably easier for me than for those without much foundation in programming. Personally, I think MatLab toolboxes should ship without all the fancy GUI crap, and be cheaper for it. I just want the functions. Bells And Whistles Toolbox should be sold separately. =)
At the end of the day, I see MatLab as being useful for research work or prototyping. Once an application grows to a certain size, using MatLab becomes impractical and counter-productive.
@Geoff: US $?
Why does Matlab get counter-productive from which certain size on?
Matlab is indeed expensive, but it is not the initial price that irritates me per se. What is inexcusable is that Mathworks do not permit the purchase of additional toolboxes unless you have an up-to-date maintenance subscription (20% of total cost per annum). Should I decide that in 2 years I want to purchase the statistics toolbox, for example, I must start again from scratch if I haven't paid maintenance. I understand that any company needs money to develop their existing software, but I see no reason why they cannot allow me to purchase the older version of a toolbox, appropriate for my licence (e.g. 2010a Stats for 2010a Matlab).
In addition, the extreme parsing of smaller and smaller toolbox functions seems unjustified. For example, why are the Wavelet and DSP toolboxes not simply part of Signal Processing?
I have recently purchased Labview to help with the data acquisition side of things and, while not perfect, NI seem to have a more fair approach to academic licensing compared to Mathworks. The software also seems less buggy. For just over £1000 I can perform real-time operations and employ stand-alone embedded controllers. To do the same in Matlab is at least twice the price.
Dont get me wrong, I think Matlab is an excellent product and I will continue using it, but they seem overly keen to bleed their customers dry. Unfortunately they can get away with this due to lack of direct competitors. It is no surprise that they are currently under investigation for anti-competitive behaviour: http://www.bloomberg.com/news/2012-03-01/mathworks-software-licenses-probed-by-eu-antitrust-regulators.html
In addition the policy to fix most of the bugs only in following releases forces users to pay the maintenance costs. There is no way to buy a stable version.
Because I switched from university to corporate pricing, Matlab seems expensive, especially if you rack up several toolboxes, but it is indispensible. The file exchange and Answers help soothe the burn, however. Other software packages are similarly costly so I don't single out Matlab.
The academic version of Matlab at about €500 is too expensive for me, a retired lecturer. Cleve Moler in his SIAM oral history interview said that Mathworks makes most of its money from industrial clients who pay the full price, so I don't think they are too concerned about indigent academics.
I believe R is a very good free alternative to Matlab, especially now that the R 2.14.xx 64-bit version gives the correct answer to sin(1e22)[= -0.8522008497671888], among other things. The 32-bit version does not give the correct answer. It has a Matlab mode (e.g., x = y instead of x <- y, etc.), and has many packages (= toolboxes) that may be loaded on-the-fly. It is slower than Matlab because it does not use the Intel or AMD Math kernels. I'm not sure if it has a JIT compiler. However it is being constantly improved.
R has become so popular that even economists and doctors are using it. Imagine! Also, R has a huge user community which is growing daily, so that help is always available, especially on StackOverflow:
Revolution Analytics has a very professional version of R that has an excellent GUI and it uses the Intel math kernel(s). Academics can get the full version free.
@Derek
Thanks for this link. I had tried and been impressed by R , but did not know the Revolution incarnation. Downloaded after reading your post. For those, like me, who want a proper GUI, it puts R in a new league. Only problem: Windows+Linux but not Mac.
The two levers for price are: Demand and Offers. I guess we all agree on a demand for professional computing/programming solutions like MatLab. So what other offers do we see? There are many FEM tools doing what some ToolBoxes do. There is R, MathCad, Mathematica, Maple (basically the symbolic toolbox) etc. But all of them have a different approach and their short commings and advantadges when compared to MatLab. Finally, you might also compare it to Visual Studio and other programming suites. If you take into account, how long dynamic memory appraoch and accordingly programmed matri-vector operations will take you, e.g. in C, and in MatLab it is just writing the math you might conclude: The price is fair.
...yes, you really need to have a functional equivalent to compare it with.
Otherwise you could compare it to a bare Linux install, a fast Internet connction, a seat with a comfortable cushion and a gallon of fresh coffee.
So evaluating the price is somewhere between those two extremes.
It appears that Lindsay Coutinho, an employee of TMW, has recently edited this question. Lindsay would you mind elaborating on what you changed and why? The potential for interference by TMW is one of the main reasons I have substaintally cut back my participation on this site.
I think the US$90 that I paid for the student edition was a steal, though I never used it. Within three months I became a research assistant and given the University site license that comes with most of the toolboxes. It's excellent. I believe there was discussion of canceling the site license due to budgetary issues (But hey, We can blow a few million dollars annually on our losing football team !!</rant>). Now that I'm pretty much through with the University and looking for a real job, I don't know if I'll bother to purchase the professional version or not. I think it would probably be worthwhile to my future employer to have me on it since automation is so easy, but who knows.
"...it would probably be worthwhile to my future employer to have me on it ..." - well that's a moot point now, since you now work at The Mathworks.
I think is too expensive. I am a student and I bought the student version and I was expecting to get everything I need after paying for it then I realized I have only few toolboxes and when I need other toolbox have to go all the way to my uni to get it. Even in my university each department have only selected toolboxes and dont have the others that they expect students wouldn't need it. Like Computer Science they don't have the communication toolbox while Electrical department have it, and Electrical department dont have Image processing while Computer Science have it.. I think its even too expensive that universities can't afford to buy the whole package !!!
If you do not make money from it you might be eligible for Home license starting at $US 150
It is even more expensive for individuals who do not make money out of using it but just to learn new knowledge. I am an engineer, and I don't see hope of Mathworks in the long run. They either reduce the price, make their team smaller and keep a good budget on their own, or they die in the competition with Python or other open source tools. I have decided to turn from Matlab to python, as I don't think thousands dollars for a perpetual individual license is worth it. I would like to remind Mathworks what Matlab was originally created for and who created it.
The stories I could tell, of saving up the pennies in one area so that another area could buy an expensive item...
I work at a university, but I have to pay for MATLAB out of my grants. MATLAB is one of two non-free (both as in beer and speech) pieces of software I use. The other is a sip/voip phone which I just haven't gotten around to replacing with a FOSS alternative. I do not think SciLab and Octave are viable alternatives to MATLAB for me. I think Python is a viable alternative for me. There is not much that I can do in MATLAB that I cannot do in Python. There is no question that if I was just starting out I would use Python instead of MATLAB. MATLAB, however, is cheap for me compared to the cost of porting my existing code to Python.
I really wish Mathworks would separate the MATLAB IDE from the MATLAB language and then make the MATLAB language free, as in speech, while keeping the IDE proprietary. I would be more likely to continue to pay software maintenance and software support in this case and not switch to Python.
Well, if you want to use Matlab, someone needs to maintain the system on which it is to be used, so there's a fair point in this discussion. And the degree to which the system is locked-down affects the various toolkits that are sold to run on top of the base Matlab. So to be brief, I think that everyone who runs Matlab needs some real knowledge of how to secure and maintain their system, if that is to be done for a remote system then they need to learn how do it remotely. If an owner/user cannot and will not take the time to learn and to keep up on effective system-administration they need to at least find someone to do it for them. This is a huge issue that certainly drives the Windows-to-Linux migration. Arguably both can be installed, the installation secured and then the secured installation maintained, and cloned along the way. The question is one of quality and robustness. I have a new Asus PN50, it takes 64GB of ram, any of the late-model Ryzen processors, has a variety of USB and HDMI ports, takes a mix of different SSDs internally and has a micro-SDXC card-reader on the front all in a case the size of a stack of 5 CD-rom cases. Swapping storage on and off it is not the problem. You still need to know what you're doing in terms of system administration unless you want to find and pay for software or an online service that can administer your system properly for you. And that has to be done, and well-maintained, before users can focus on software-development using that system. It's an argument for Azure or Matlab Online or some similar platform, accessed through a VM or Chromebook style front-end, and a completely secure local boot device.
It's nice to be able to ignore all that and just have users click on the Matlab icon and be happy.
Just as it is nice to ignore the question of why someone wants to run Matlab code without buying a Matlab license. Well it's not just running the code, but running it on the proper platform. Gotta draw the line somewhere. That's why Google came out with the whole Chromebook platform and Google online apps. Mathworks could do something similar that runs on Chrome or Firefox, then develop their own Linux distro to support that, or their own version of Azure...the question still comes down to how to secure the platform and who will assume responsibiity for that. Who really wants to take on that responsibility?
I also feel the tone of the discussion has become so bitter I don't feel I should participate. Although that might just be me misreading the situation.
I feel this discussion about Microsoft Windows installs has veered quite a bit off topic.
Let me see if I understand your proposal correctly. Do I understand correctly that I should:
- buy two new computers that are duplicates of each other
- install Windows here
- ship one of them 1000 miles to my mother
- wait until one of my mother's Elder-care volunteers with computer skills wanders by my mother's place to plug the computer in to all the right cables (my mother is not physically able to do the necessary bending anymore)
- each time a significant Windows bug fix or update comes along, install it first on my master copy, then clone the installation and ship the cloned system to my mother
- wait until one of my mother's Elder-care volunteers with computer skills wanders by my mother's place to do whatever is physically necessary for the installation of the cloned system
Is that correct?? And this would represent "serious IT skills" but downloading the update to her system using a remote control program over the network would not represent "serious IT skills" ?
So... a question then: those people "with serious IT skills": where do they get the "reference Windows install" without installing locally over over the network ?
No one with serious IT skills would install Windows from source either locally or over a network.
They'd clone a reference Windows install and activate it, preferably using a site-license.
...Yet another reason that many prefer Linux.
"Your reply reveals your true problem: you have yet to even think of a good solution for that problem."
What "that problem" refers to is not clear?
Because if you were referring to my remark about Windows being slow, then I would take note that you do not know what hardware I am using, or my reasons for using Windodws, and that you have not been CC'd on my discussions with hardware and software manufacturers.
You're misstating the problem.
The problem is that you lack the ability to use Google in imaginative and productive ways.
There is absolutely no way I can solve the problem that Windows takes hours to install, and is very frustrating to manage.
Your reply reveals your true problem: you have yet to even think of a good solution for that problem.
On the other hand I have to admit that it's hard to see someone adding significant value to a problem in 2022 who has "little to know knowledge of programming at all" or for whom that is a real hindrance to their adding significant value to a problem.
What we're mixing in here is the idea that someone WOULD find it difficult to do valuable work without a tool unlike Matlab and without extensive computer skills. To accomplish a task, we must develop both the skills and the knowledge required and do the work required. Usually this is somewhat circular. This differentiates humans in many ways, but only one group actually does the work. The rest sit around babbling about honing their skills and acquiring and refining knowedge but don't get the work done.
"You could on the other hand sit down in front of a Windows machine and run Excel and write VBA and get a lot more done a lot more quickly with little to no knowledge of programming at all"
... but certainly not within one hour! It takes me several hours each time I have to install Windows -- and I am on a relatively fast internet connection.
um, that's neither true nor relevant. There are 500 accessories installed for free on a Windows platform. If you don't need them and have no real use for them then why would you care? They are installed there, and often made un-removable, by MS to add value to Windows as an OS and as a work platform. That is what you're paying for when you buy a Windows license, that is what keeps Windows relevant. Bundling is a common practice for this reason.
When it comes to Matlab, Python and C/C++, which are specific tools, if you don't know how to write code in any of them, why would you care whether they are there "when you first boot up" or if they have to be installed or not? You're confusing a historical interest with a technical interest.
What I do see is relevant is the fact that in 2022 you can certainy install Linux (I use Ubuntu) for absolutely free onto an x86/64 platform and have it up and running inside an hour with Python and C++ on it. Does that mean that you can just sit down in front of it and start to write Python and C++ code...if you don't know anything about either one? You could on the other hand sit down in front of a Windows machine and run Excel and write VBA and get a lot more done a lot more quickly with little to no knowledge of programming at all, starting with scripting GUI actions, because the support environment is there and well-developed. With OS you have got to build a support environment or find one to install and hope that it works properly. This is where the "you" in "there's a lot that you can do in OS for free" comes into play. You still need to know what you're doing, well enough, broadly enough, to make your time and effort profitable, no matter what platform you work on and what tools you work with. And that is a major factor when it comes to paying for commercial licenses. You either have put the time and effort into OS to make those payments redundant, or you have not. If you have, then this entire discussion is preaching to the choir.
If you have not, then you either buy or access a suitable license and continue your work or you develop your OS tool skills or even both. But you don't just abandon Windows and commercial licensing and jump into Linux and OS development with both feet. Any more than a mechanic tries to do all of their work with a Phillips screwdriver. You develop a tunable migration plan and exploit it as you gain the technical ability to do so.
I just don't understand the logic of trying to convince Matlab to provide more support to OS and to Python specifically. If you want to use a sledgehammer to drive nails, be my guest. I'd be more interested in expanding OS and Python to better interface with Matlab but then you'd run into problems with their maintainers and maybe that explains the question best of all. To me it's simple. Call the Matlab engine from within your OS tool and run your Matlab code in the Matlab engine. All that you have to do is figure-out how to interact with the Matlab engine. Just as all that you have to do is figure out how to access a valid Matlab license. Porting Matlab code out of Matlab into a third-party interpreter is just...crazy. Amazingly Matlab does actually support this with the Coder tool...it also supports COM Automation, and it also supports packet-level data-exchange. Do some research into the capabilities of what you've already paid for, you might be amazed.
Matlab always has the option of developing a configurable OS platform for clients who make end-use driven purchases. Python fanbois have that option as well, plus the choice of offering and supporting such platforms for free.
It amazes me how Python has spread to so many platforms, along with extensive math libraries - and it's already there when you first boot up. C++ is the bread and butter of many programmers, yet you generally have to install a compiler before you use it.
MATLAB is not cheap. If you need 4 toolboxes and a commercial license, you have to pay some thousands dollars. Using Matlab for scientific purposes costs about the half. (For my argument a factor of 2 or 4 does not matter...)
In a real business case, a customer can invest X dollars to solve a specific problem. If X is smaller than the costs of MATLAB, MATLAB is expensive.
If X is about 10 times the costs of MATLAB, MATLAB is getting very cheap, because its very powerful and well tested toolboxes allows a rapid prototyping, implementation and testing of the program.
So cheap or expensive is a simple balance between the total costs for development and the utility value of the program measured in dollars. If I compare MATLAB with a low level langauge as C, it is cheap, even when a free C-compiler is used: The previous question I've answered here concerned POLYFIT. I suggested to create the Vandermonde matrix manually and calculate R\(Q*y) directly - imagine the effort, if I want to explain this based on C99 and BLAS/LAPACK, including checks of inputs and an automatic usage for DOUBLEs and FLOATs...
A comparison with SciLab, Octave and SciPy is more reasonable. The student version of MATLAB (about 100 Euro) is expensive, because I do not see anything a student can learn with MATLAB, which cannot be learned with Octave. In opposite to this, 5000 Euro for a scientific license for MATLAB and some toolboxes is cheap for the development of a large program (>100.000 lines of code) for clinical decision making.
I think Samuel and Mike have the right way of thinking about it.
In university, I have 7 years experience with MATLAB. But I recently graduated, so I would have less access to my license. I was thinking of switching to Python, numpy, and scikit learn to get into machine learning/deep learning. But these packages aren't seamless and have noticable quirks coming out of a MATLAB environement. So I would basically have to re-learn a lot of things from scratch.
My expected salary for a new job in my field would be in the range of ~100-120k USD per year. A perpetual matlab liscense is ~$2500 (no packages?). So if Python takes me more than a week or so of study to learn (a week salary is $1923-2307), it is already more expensive than MATLAB in that sense. And on top of that with MATLAB you get seamless integration of different packages, continual software improvements with quallity assuance, and teams of experts who want to answer your questions.
So from this perspective it is easy to see why MATLAB has remained competive for many applications. And I think this kind of buisness model for MathWorks is similar to the Microsoft product suites. If everyone is very knoweldgable about their programs (from learning in elementary through high school, for example), the cost of switching to new software becomes really high, and the program remains valuable even if other free programs are available (for example the same reason that Windows dominates over Linux).
"MATLAB is not cheap. If you need 4 toolboxes and a commercial license, you have to pay some thousands dollars. Using Matlab for scientific purposes costs about the half. (For my argument a factor of 2 or 4 does not matter...)"
At $1k per license and 4 licenses, that's 1 month salary for someone earning $80k/ year before taxes.
is it...the price of an Android game? No. But that game could have billions of people spending hundres of billions of hours per year using it. We live in a world where hundreds of millions of people spend billions of hours each year playing video games and surfing the Internet as well as real money on stupid crap like buying and building expensive cars and houses and drugs INSTEAD OF developing real value using Matlab.
Don't care so much about the cosft of the software while ignoring your opportunity to make money with it and the way that you waste money and time in so many other ways.
I totally agree with Mike. When comparing Python to Matlab, people never consider the time they waste dealing with the bugs in Python. Time is money and Matlab is saving your time.
Matlab is not expensive if it saves you time over the long run. I learned Matlab in the 1990's when there was no Octave nor Python. I have since learned Python too and I think Matlab is easier than Python. I think Matlab's built-in library functions are more trust-worthy than Pythons. Matlab is just easier, especially if you want to create fantastic looking plots. I use Matlab most days, hundreds of days each year. Some days it might save me 10 minutes, some days it might save me hours, but in the long run it pays for itself.
I think today you need to compare ti to Pyton and Julia. I think it would put MATLAB in a very bad position.
"So cheap or expensive is a simple balance between the total costs for development and the utility value of the program measured in dollars." Am I correct? This means that pricing based not on cost bases but on revenue of users. Like is steel manufacturers sell steel for car manufacturers based on one price and to builder for their price based on their utility cost. Personally it seems to me that Matlab costs is not transparent at all. I always should ask a cost for every thing. Provided prices are valid only 30 days. It brings feeling that manager is trying to get as much money as it can. It is like pricing in monopoly and oligopoly economics.
@Walter: I've heared, that Octave is a calculator only, as MATLAB ;-)
Graphics are supported, see e.g.:
http://www.gnu.org/software/octave/doc/interpreter/High_002dLevel-Plotting.html#High_002dLevel-Plotting
Has the graphics of Octave improved? They were nearly non-existent when I last looked at it.
I am of mixed opinion on this.
If you have professional work to do, you use professional-quality tools, and it is normal to expect them to be higher priced, as they generally pay for themselves in generality or durability or in being designed for what you as a specialist need. Those traits cost money to make and people are willing to pay appropriately.
From that perspective, perhaps some particular toolboxes might not be worth the value, but overall I do not think MATLAB is unreasonably priced.
However, there are people in the MATLAB community who donate a lot of time and effort to helping others, and those people often get little or no pay for those efforts. To those people (especially the retirees), there is little realistic hope of earning back the cost of toolboxes or of keeping MATLAB reasonably up to date: to those people, MATLAB is expensive.
I do not have a solution to offer to this difficulty at this time.
Benjamin, where would you get the MVP's sufficient that all questions were dealt with "way quicker than here" and with high quality answers? Have you estimated how many MVP would be required?
If the MVP principle is used that independent (non-employees) only are to be considered, then that would eliminate a number of the people who often answer questions here and in the newsgroup. Please keep that in mind when you think about individuals who might be nominated.
I have it in mind that any structure like that would rely on _individuals_. I do not think it appropriate to name names in public, but I ask myself, "Okay, whom?". I know of some people who would probably decline for various reasons.
I also suspect, Benjamin, that you are significantly underestimating the portion of the questions here that are not really programming questions, but are instead science or mathematics or technology. MATLAB is "The language of technical computing", and MathWorks is "Accelerating the pace of engineering and science". If it has to do with engineering, science, mathematics, or technology, then the question might be posed here.
It is not uncommon for questions here to be about some focused aspect of a topic that has literally been researched for decades. Those questions are often posed by someone who has no background in the topic and has not done research in the topic nor read the relevant papers that focus on that specific aspect. "Complete code" is an oft-expressed desire. Where are you going to find the stable of volunteer scientists and engineers ready to promptly handle whatever question comes up ?
I consider the Microsoft MVP program also a very good idea. I had some time ago to deal with MS Access and my questions where answered way quicker than e.g. here and the quality of the answers was commonly high. Most of the responding persones where Microsoft MVPs. I would be willing to pay a higher price for MATLAB, when this kind of support (which is cheaper than setting up a correspdonding amount of support engineers) is being established.
I love the idea of license per participation but I would appreaciate more if TMW had something like the MVP (Microsoft).
@Daniel, @et al: Please excuse my brute abbreviation of the list of famous CSSMers and Answerers! I just picked 4 frequent users from the Google-Groups overview of CSSM. I think, Yair should get a version *without* documentation ;-)
@Walter: I've bought a more powerful computer to cope with the keyboard latency in this forum. A cheap Core2Duo with Win7 is sufficient for streaming of HD videos and typing answers here... An update of the preview would have been more appealing.
It amazes me that Mathworks doesn't do this. The et al. in Jan's comment is short. You are talking maybe a dozen licenses. Mathworks could make the licenses expire after 1 year and make renewal dependent on reputation points, CSSM posts, or FEX submissions.
Dear Walter, the solution is simple: TMW could offer a free MATLAB+3 toolboxes academic license for 3000 reputation points. If I calculate the costs saved for the technical support, this would be a fair price. I think it should be possible to find a solution for Steven, Bruno, ImageAnalyst and Matt J, et al.
Sign in to participate
