@P.Fonseca I suspect that if Flash player cost $1 then it would be universally considered as overpriced trash. By giving it away for free Adobe has created a captive market for their other applications. Very clever move.
@P.Fonseca so this is where I am unclear about the sort of product you are seeking. Okay we have the OEM model but lets put that to one side. For a Flash comparison I don't really see why you don't like CDF enterprise . If I as a developer build something and "sign" it with CDF enterprise then we have the parallel with Flash: Flash = user downloads free Flash player. CDF = user downloads free CDF player.
Many Flash based web apps are data presentation which I've seen adequately done with javascript libraries. I have recently spoken to people about CDF for market data display but one built a Flash app the other went with javascript. In both cases this was because 99% of what is in Mma is not needed for a lot of these sorts of data presentation apps (since we are talking about Flash comparisons).
What would be nice is an iPad version. People can still build javascipt apps but Flash won't work on iPad. I was browsing Mike Croucher's site yesterday and learned that Maple have had an iPad "version" for over a year. The problem with being a feature follower, which IMO Wolfram mainly is (exception Wolfram alpha), is that people get locked in to usage of other products and then need a compelling reason to switch.
@OleksandrR. I'm not talking of Flash as a comparison technology/application with Mathematica. I was comparing business models. Concerning the CDF enterprise: CDF only allows me to do applications tailored for CDF interface.
Once again, Mathematica is a technical product, and industry doesn't want another interface. We want to integrate it with our working environments (whatever they are). Everything else is a toy. That means calling the Player (or an interface free runtime) from exterior apps, and retrieving results. And this isn't existing for the masses. It only exists with special contracts.
And since no one knows what Mathematica is, it becomes to cumbersome to even try to use WRI technology
@P.Fonseca yes, I agree with you. Perhaps WRI is philosophically opposed to that sort of thing, though. I was amazed that Leonid's RLink appeared officially with Mathematica. Up until now WRI has pretended as if other similar software simply does not exist.
@OleksandrR. not only that it pretended it had the best statistical software on the market based solely on the fact that it supported the largest number of built in distributions. But for those who view Mma as a prototyping tool or technology integrating tool it is a nice addition i guess (something that many users have wanted so nice that users got listened to on that one)
Once again, what would be the impact of a free runtime, instead of the existing very limited RTL? Would I start to use Mathematica in all my projects? Most likely. Would that give visibility to WRI? Obviously. Would my fellow colleagues gain interest in Mathematica? Most likely. Would that mean more Mathematica licenses sold? Seems logical. And again, we are at the Flash methafor.
@P.Fonseca still not seeing the business model differences. Flash: you buy whatever software is needed to develop a Flash app. You make a Flash app. Your users download a free Flash player and run your app. CDF: you buy whatever software is needed to develop a CDF app (CDF enterprise). You make a CDF app. Your users downloaded a free CDF player and run your app.
@MikeHoneychurch the Flash player is free and can do anything any Flash content creator wishes. The CDF player is free but incredibly limited (like MS PowerPoint Viewer or whatever); to do anything nontrivial requires either a paid-for version (which is not that cheap) or in the worst case a special contract with WRI.
@MikeHoneychurch a CDF app is a toy in most industries. Lets suppose I need some kind of MS visio interface (block diagram as the one of system modeler). How can I do it professionally with Mathematica dynamics? What I can do is use a program that manages it correctly, and link it with mathematica for the needed analysis. But I can't link libraries to the Player. (the block diagram interface is just an example...)
The key questions is: how many Player pro licenses were sold in the last 4 years when compared with the extra Mathematica licenses that could have been sold with the extra exposure of WRI if every single one of us, Mathematica users, would have integrated Mathematica into everything (since we simply could)
@OleksandrR. that is incorrect. Yes the CDF player is free but its limitations depends only on what it is being given to run. A CDF created by a Mma user has a lot of limitations and is only for toy usage. A CDF created by CDF enterprise can supposedly do pretty much anything when played with the free player. I have a few things to do before I go to lunch but I have copies of CDF enterprise, Player Pro and Mma so might make some comparison later today.
So I think there might be some confusion going on here? Apps created by CDF enterprise remove the need to have Player Pro ...they make the FREE player act like Player Pro. That is the whole point of the product.
...or to put it another way, what functionality do you guys believe is excluded from free CDF players user who are using an app created with CDF enterprise?
@MikeHoneychurch oh, really? I haven't been paying attention to developments regarding the Player since I can't really see myself ever needing it for anything.
@OleksandrR. yes. I have asked a lot about this over many months due to some projects I have worked on. In that time things have changed a bit but it is my understanding that that is the current model/usage. So rather than sell x number of apps that require x* player pro licenses you buy a CDF enterprise and "pro sign" the CDFs. Your users can then access full functionality even with the free player
@MikeHoneychurch I don't think there's a confusion. Can I now call a runtime from my company very specific tools? Or will I only be able to use the CDF interface? I think I understood correctly that we can now create CDF that have the player pro capacities, even running on the free player, if they are built with Mathematica enterprise edition. If this is correct, there's no confusion here.
If it's really the case that CDF Enterprise can make the player do anything that one can do with Mathematica, I think that fact is not very widely known. If you can show a test case of plugging in a LibraryFunction or a MathLink executable, or doing something with J/Link, I think that would be good to have on this site.
@P.Fonseca but this is where we are apples and oranges. Can you do that with Flash? We tend to be mixing discussions of the Flash business model with your specific needs in your company.
@OleksandrR. It allows the free player to do what Player Pro can do. If Player Pro can do those things you mentioned then someone using the free Player could do that too. ...I am going for lunch but maybe we can continue this another time.
@rcollyer I think we can both be thankful we're not Portugese... 200 page thesis, 10(!) publications (according to my Portugese colleague), and a public defence with 5 or 6 external examiners seems to be mandatory.
How can I put it in a different way... Do you all think the MATLAB runtime is a completely absurd thing? The simplest example: I want to integrate the energy loss along a flow passing through a channel. This is made with 4 lines of NDSolve. I compile it and then I call it from my specific company software (a spreadsheet that is not Excel, so to avoid the comment that I should get Excel link).
Since it uses NDSolve, I can only make it work as long as I have Mathematica installed, and even that risks of being more complicated than it should (than it is in MATLAB). I want to give it to a colleague. How can Mathematica enterprise help me? To my knowledge, it can't.
There's nothing I can do with the exception of reprogramming it in another language (or buying another version of Mathematica for my colleague, or enter in contact with WRI, and start an OEM partnership, for 4 lines of code...). How many lines of code justify the work? IMHO, too many in the majority of the cases.
@P.Fonseca well, now that I know the stringent requirements I will surely look at Portugese academics with a new respect! (And seek them out for collaborations...)
@P.Fonseca worth noting I suppose that the site licence can give dramatic discounts. For instance I got 90% off by using my university's site licence.
@OleksandrR. specially because we are not expensive (again, thinking on enterprise environment)... Up until around 2006, we could only be called Msc with 5+2 years of training (beyond the typical 18 year old). Now, to put us in an equivalent scenario as the rest of the world, we have downgraded requirements...
@OleksandrR. WRI knows how the academic world works, but taking by the discussions I had before, the production enterprise or industry (that is, anything beyond institutes and R&D centers) is somehow a mystery.
@P.Fonseca I suppose you have to ask, what is Wolfram's motivation? It seems he has no great desire to become rich, but wants to be well-regarded among academics.
@OleksandrR. I agree, but wouldn't selling more licenses help on the tool development. If you look at their site, you will see lots of mentions to other targets than pure academics
@acl I am not sure how seriously he thinks people take that. He is certainly more intelligent than I am, so I doubt he is totally in the dark as far as how people view his academic contributions. Although, I do detect a bit of egotism with the recent trend toward calling every new feature in Mathematica the "Wolfram ..." rather than "Mathematica ..." even when it has no conceivable application outside Mathematica.
@Kiyura In mathematica, expressions are trees. Usually, you define rules associated with the head of the tree, but TagSet lets you associate rules with parts below that, but in mathematica only 1 level deep
@OleksandrR. so? to read it, nobody thought of CAs except Wolfram. In reality, there's papers by others, before his, with thousands of citations. This is just one example
@OleksandrR. Well by default, my defense is public at least partially. Then, my committee kicks everyone else out and grills me for a while, and finally kicks me out. Once they've laughed at my expense, for a bit, they will come get me and tell me the good news.
@OleksandrR. indeed, maybe one day he woke up and exclaimed: "oh no! other people have also been trying to think about life, the universe, and everything as well! they may even have had original thoughts!! How could I have missed that!"
@acl perhaps he didn't read the literature seriously before writing NKS. Then, he might have had a shock when he was criticized for that after it was published, and realised that the field was a lot more rigorous than he had given it credit for.
@Kiyura have you tried to contact the author (dbwagner@princon.com). Since there is nothing on google on the recent years, I don't know if he still answers... I could probably lend it (where are you in the world?)
Just saw it... United States... The other side...
e-mail me if you want to arrange something (but I think that a library exchange program would be easier)
@rcollyer ah, okay. Well, not sure if that is good or bad per se. I have two examiners: don't know the external examiner at all (usually suggested by the supervisor, but in this case my supervisor didn't know him because the change of project called for someone from a different field), but I know the internal examiner well and trust his judgment.
@P.Fonseca as I said before a left the chat a few hours ago I think we are talking past each other. I see nothing substantially different in the CDF Enterprise/CDF Player business model than the Flash/Flash Player business model. Your parallel discussion deals with usage cases that fall in the gap between large scale OEM deployment as building apps for CDF deployment. Yes I agree that is probably a gap but also would expect that the returns for filling that gap would greater exceed costs...
commercial usage of Mma is very very small
correction: gap between "large scale OEM deployment AND building apps for CDF deployment"
I'm trying to use Mathematica as a serious tool inside the company I work (I'm physicist and work for a Supermarket Group as Data Science Manager). I have the same terrible problem with distribution told by @P.Fonseca. I bought the enterprise edition and want to see what can be done. They told me that I could create my own signed CDF with MMA9, but I had not succes in do it yet.
@rm-rf, I am not, but I live near my alma mater. Maybe I could still get a loan in as an alumnus
@P.Fonseca, Oh, thanks for the thought, but it's not worth the effort and certainly not worth the shipping. Have you read the chapter m_goldberg mentioned? Is it that great of a resource on the evaluation?
@Kiyura what chapter was that? I don't have the book here with me, but I can look into it tomorrow
@Murta I don't see a problem with your scenario. As long as you can create the CDF app, you can, with the enterprise version (the one that costs around three times Mathematica pro), create a CDF app that can be run on all players (free or pro). At least, that is my understanding.
@MikeHoneychurch to my knowledge, that gap is HUGE. And this is the essential difference of understanding. My company division where I work has around 2000 people. It is neither a manufacturing company, nor an essentially service company, meaning that a great percentage are engineers or from other technical origins. Nevertheless, very few have heard of Mathematica (1/25), and most have heard of alternatives (9/10).
There is only one license of Mathematica (mine), that already toke some time to justify (why should we have this thing that nobody has heard off). Every time there's the need to develop a specific functionality into existing tools, I guarantee you that everything else is more interesting than Mathematica (should we use standard libraries from the market, free or known, or enter in discussion with a company that no one has heard off...).
While there is the gap between large scale OEM, and the limited CDF use, few will know of Mathematica, and this will effect the amount of "large scale" companies looking into Mathematica as a solution partner, and the amount of companies wiling to pay 9 k€ for the signing of CDF pro apps. Now imagine that my company uses Excel (and although no one here understands why, this is the pure reality of a HUGE part of the world). Imagine I want to add a function into Excel.
I can program it entirely in VBA, python, or .NET, or link an existing library to it, programming just the communication interface (if the library doesn't have it yet). Now I want to do it with Mathematica. Either a Mathematica copy for each user (or a license server for a certain number of licenses of that thing that nobody has heard off...), is installed, or not possible.
If only the compiled library could call the player kernel... Suddenly all computers in the company would have the player installed, and I would have done in one day what WRI is trying for years; to get it used. Since the kernel would already be installed, why search for something different next time?
Just buy Mathematica licenses for the other interested developers, being it for simple CDF tools, or integrated needs. Nothing goes against the OEM stuff, since outside company distribution can be limited by contract (as it was the case before).
Imagine that every time the Player Kernel is started during a OS session (called from an external library), a small banner appears at the right bottom corner of the screen (thinking like Windows, don't know the appropriate corner for Mac or Linux ;-), saying "powered by Mathematica Kernel; Wolfram Research, your technology partner". HUGE! no need to thank me ;-)
And all this success because of 4 lines of Mathematica code that I was lazy enough to transform to 20 lines of VBA.
"What is that technology kernel stuff that is already installed everywhere? You know, the thing that appears each time we call program x. Can we use it too for this other case? Who knows how to use it? Ok, I will talk with Fonseca for him to explain me how he did it."
And there goes some more Mathematica licenses here and there.
fyi, someone at student forum just posted a bug in Integrate, I verified it is a bug. Here it is Integrate[Cos[x]/Sin[x]^3, x] the correct answer should be -1/(2 Sin[x]^2). I send bug report to WRI just in case.
@P.Fonseca, I am student. Reason I use Mathematica is because I can make simulations which helps me learn something better, in it much easier than anything else out there. Also, One can run this on the web. There is nothing at all right now in the market close to this. Applets are pain to do. I can do a GUI and simulation in 1 hr in what it would take one a day to do in other system, even with Matlab (which I know very well also)
@NasserM.Abbasi but the corporate world doesn't have the same logic as the academic world (I've been at it for quite some time now). It has different needs, opinions, and visions than the academic world (or institutes or R&D). And it is this world that produces.
Even if it is the most logical thing just to use Mathematica, even if that means some trouble with license servers or some special contract with Wolfram, your enterprise "ecosystem" will not start using just because it is the best thing.
After some years on the two worlds, I know that this is hard for each side to understand. I also use Mathematica everyday, and I'm very happy with it. But I use it because I kept a link with the academic world. On the corporate world, things need more integration than on the academic environment.
There are special tailored tools to do very specific tasks. The CDF interface is too disconnected. Functionality gets build on top of existing tools. I don't mean that there aren't cases where CDF is applicable, but I can say that this is definitely not the main stream. To present reports with CDF when there's no other reason to install the Player besides the reports, is a no go.
Why should we loose time maintaining a new app, just because of a report. If it drives a company core business, and gets integrated into the technology stack, then this is another thing.
(please take no offence with the "that produces". I mean "that directly produces money", or "that sells").
Some people don't understand what means to maintaining a free player (I had several discussions on the subject). It's free, why should it be difficult to install in whatever computer? Because every tool needs to be tested internally (you know... to avoid having half a company stopped because all computers are formatted in the same way, and suddenly all suffered from the BSOD...).
This means that asking for it to be installed for a CDF toy (sorry but I insist) or to have a futuristic report is a no go. It needs to enter into the core business. It needs to really generate money.
@P.Fonseca, I used to work also. I left to go back to school, so now I am student. in my last job, I used to write Matlab GUI apps that many others in the company used to do scientific analysis on data. I would compile it to .exe, they would then download the app and run. I used Matlab GUIDE to make the apps. Mathematica at the time did not have CDF's. I can see now that I could have used CDF's for the same thing, and it would have been much simpler actually to do.
I can't comment on licenses and such issues. But from technical point of view, CDF's makes it very easy to do things that is very hard to do with any other technology now. Use Flash to do what I do now with Manipulate?? Java Applets? Write 1000 lines to do what I can do in 20 lines?
@NasserM.Abbasi I insist on the same thing. Can you easily integrate Mathematica technology on the corporate already existing stack of tools? Only if you do completely independent tools, with their own interface (only acceptable for few applications) or if you buy full mathematica licensees, and a license server, or if you do special contracts with WRI, etc. Too much trouble to get started, so, no start. How to call it from a CAD system, or from a block diagram interface tool, or, etc.
@P.Fonseca, I do not think there is a need for this deep integration for a commercial product to be success. Matlab makes millions of dollars because it is used by many to design engineering applications (large companies use it every day). Simulink is used to design many things everyday. And it closed as well. If WRI can just make a real compiler (i.e. Mathematica code-> .exe) then it will be all set.
That's what I'm saying! The only thing we need is the same integration MATLAB has, and as you said, they make millions of dollars. But probably they are wrong having a compiler connected into a free complete runtime...
We have Mathematica code -> .exe but it only works for simple stuff. So a little pointless for integration.
And all this discussion is about the need to it to compile to complete useful libraries, that access the entire kernel of Mathematica, without having to produce special contract with WRI, exactly as MATLAB is doing, and making millions of dollars. Everything is existing, except the vision...
@P.Fonseca, well, WRI has the CDF. Mathworks does not. Mathworks has a compiler, WRI does not. (I mean a real compiler, to make .exe directly from all Mathematica code). So it is draw right now. For me, I find CDF more useful than a compiler. (I'd rather run something on the web, than having to download it each time). But for companies, I can see that having an actual compiler would be much more useful as you say.
Thank you. That's why WRI is on the academic world, and on corporate world, only MATLAB is known. But probably WRI makes on purpose to miss more than half of the market. For MATLAB to have CDF, they need to create the tools. For Mathematica to have compiled libraries that access the free CDF player kernel (free as MATLAB does), they just need to decide it, because they have everything they need. So where you see a draw, I see a premeditated decision to be behind. (being harsh)
@P.Fonseca, fyi in Matlab 2012b , Mathworks added now a new feature, called 'apps', where with one click of a button, a single file with extension called .mlappinstall is created that includes all files needed to run the main matlab file. You'd still need Matlab on the other end to run it. But it makes packaging and installation and moving apps around much simpler than before. Here is a link mathworks.com/discovery/matlab-apps.html
@P.Fonseca, yes ofcourse. One needs Matlab to run the new matlab 'apps'. That is why I am saying is that Mathematica CDF are much better. One does not need Mathematica to run them. I know CDF's are limited in what they can do, but with this limitation, one can still do allot with them. If WRI would allow InputField[] for arbitrary input in the standard CDF version, then that will allow many more things to be done also.
But still on a non integrated tool. The world searches for integrated solutions. That's why we need a full compiler solution (as in Matlab with the free runtime)
@P.Fonseca Yes, I have the Enterprise, but the problem in the Corporate world is exactly what you said "It's a different logic". For example, what I really would like to do today, was some way to integrate the power of Mathematica calculation in my Purchase Spreadsheet (Excel) in a simple way. This spreadsheet is used by 20 persons. Today, the only way to do that would be to buy 20 Mathematica licenses, to execute a very small quantity of code! It's not viable!
Sometimes in the Wolfram Conference this year, talking with some insiders, I fell that it's hard to explain this for then. They have unlimited license access!...
But still can't resist that I have similar scenarios (with same quantity of end users), and that doesn't justify custom cases from WRI.
No. They can't understand. It is another world. I've tried several times. So describe it to us, and we will think on a custom solution: it difficult to implement standard solutions that are not known; it is impossible to implement custom solutions for this size... No one will be interested on the investment. If there is no small applications, there will never be a jump for bigger ones. Typical comments: but why Excel? But why not a custom CDF interface? It is explaining from too far away.
What really bothers me is that I can't see the Mathematica platform growing in the corporate world in the current model. It will be restricted to a few crazy people like me and @P.Fonseca who came from academic world. The site examples are unreal. Who will learn Mathematica to do reports!
@LeonidShifrin my last 24 hours have been dedicated to trying to explain why WRI can't get into the corporate world. I would classify my success at 30%...
from @P.Fonseca "Imagine that every time the Player Kernel is started during a OS session (called from an external library), a small banner appears at the right bottom corner of the screen (thinking like Windows, don't know the appropriate corner for Mac or Linux ;-), saying "powered by Mathematica Kernel; Wolfram Research, your technology partner". HUGE! no need to thank me ;-)"
@Murta I see. Hope you won't have to wait for long. What I totally agree with is that there should be reasonable ways to connect to the kernel from other tools
It would be HUGE! You install the player, and you have everything. A calculation kernel for the apps your developers prepare you. A training/presentation interface. An interface for some small power tools (the CDF stuff).
@P.Fonseca It is not realistic to demand a free M runtime, because then it can be abused easily to not buy M itself. Perhaps at some point in the future that might be more real, I don't know. But definitely not now
@LeonidShifrin It would only the accessible by the Mathematica compiled libraries.
Of course there's probably some things needed to avoid programming with Mathematica access to all functions, in a kind of Front end replica. Put that is already a concern in the Players.
@Murta Yes, I also think the same way. But you know, I went through a very long stage of thinking that M should be central (including the interface) to all systems that use it. This was when I was in academia. When I went to industry, I relised very clearly that the first thing to do is to support existing workflows and integrate M with them. No one in the industry (except perhaps startups) would change their workflows in any significant way without knowing for sure what are the benefits
Thank you @LeonidShifrin. I've been discussing exactly what you said: "I relised very clearly that the first thing to do is to support existing workflows and integrate M with them. No one in the industry (except perhaps startups) would change their workflows in any significant way without knowing for sure what are the benefits"
@P.Fonseca Because Matlab has a different model, it is built on toolboxes. So, their main value (profit) is there. I don't know whether one can use toolboxeswith their free runtime free of charge, but I guess the answer is no
@Murta Well, RLink is not open source (although the source is fully accessible), and it is not "mine" - I wrote it, but that was an assignment I got from WRI, and of course I got paid for that
@asim I think advertising on RBloggers is a good idea, thanks!
@P.Fonseca Yes, I agree. But for M,it is harder to separate that core, due to the way M is built.
@LeonidShifrin I came with the Flash analogy (although I'm almost sorry I did), that if the Flash kernel cost 1$, no one would have used it. Since it is free, Adobe made huge profit on the builder and creative suits.
But I can live with a small price like the enterprise one (I think... economy isn't great for anyone).
@Leonid: I sometimes wonder how WRI makes decisions about what functionality to include or emphasize. It can benefit a great deal from touting its compilation to C and by making more and more of its underlying functions compilable. I almost stopped using C once, I figured out how much one can do with just Mathematica
@P.Fonseca Yes, I realized that workflows change slowly and only according to serious internal needs in a company
@P.Fonseca There are fundamental differences between Flash and Mathematica.Once Flash kernel has been developed, it does not need as much work for next versions, because it is really the kernel
@LeonidShifrin but still, if it cost 1$, flash would only mean the bright light on top of your camera. So, Adobe money came from a Builder it-self (in our case, a Mathematica version that would allow building for the kernel).
@asim I personally share your views, that's probably all I can tell you. I don't really know how the decisions are made, this happens way above my level :)
About MMA popularity, I have send this mail for TIOBE, that I would like to share here. This has all to do with this topic:
"Hi I would like to know if Wolfram Mathematica could not participate. I think that with wolfram alpha it gets a little bit more popular no? It has a very active google groups forum and exists since 1988. Tks in advance Murta"
Hi Rodrigo, Thanks for your feedback to our TIOBE index. Mathematica has been tracked for almost 10 years now. Currently Mathematica is at position 128 down from position 118 last year. In the past it was quite popular but as you can tell from the attached trend graph its popularity is declining. I hope that this answers your question. Regards, Paul
As a business school professor, we teach our students to think in terms of Customer Acquisition, Retention and Expansion. Different actions are needed for each of these. Unless functionality is introduced that brings in new users, popularity is going to stay the same.
@Murta That's pretty sad. But I amnot very surprised. In terms of programming, M remains a closed system, with the great many open systems /languages being used more and more now. And on the other side, the deployment options which aren't really sufficient. The biggest mistake for the deployment is to make M interface central to all of it. Most people would prefer to have an access to M kernel or some executable somehow produced from / by it. I don't think this has been widely understood
@P.Fonseca Yes, I think it is important. But I don't think it will change much. Changes in this direction would require serious technological changes, as well.
@P.Fonseca I'm afraid I will have to leave rather soon too.
Yes.. This is partly a positioning issue. People think it is just good only for symbolic stuff and for anything thing statistical these days R is the medium of choice.
@LeonidShifrin Have you seen the Octave - Matlab model? For example, the standford machine learn course uses Octave, that is a free version of Matlab in the bash. So, students start to learn the syntax, and it's natural to go to MatLab. There are thousand of students in all the world doing the class, so they are potential future Matlab users!
@LeonidShifrin If changes in this direction are difficult to expect, and if the world is more and more using opensource languages, this needs to be a awakening call to WRI. They have a great software, but even great things die, if nobody uses them. (I'm still here for the next couple of minutes).
@asim, @NasserM.Abbasi, @Murta , @P.Fonseca I just checked the number of books published which use Matlab, for 2012-2013(incl.pre-announced titles) - it is more than 200! Now, this is a very good indicator of popularity and usage trends
I think with Matlab, the community is more open. I mean go look at Mathworks web site. Hundreds of people participate there each day. Many different forums there, which any one can participate and post work much more easily that with WRI.
I think even Wolfram alpha was a distraction. I am not sure how much they make off it. Maybe the Siri stuff was profitable, but all those "Apps" do not look appealing.
@Murta In my opinion, no comparison. All books on M are at least few years back, quite dated. Very few modern books covering things after version 7. I talked to some people at WRI who are responsible for publishing, I think they have somewhat different views on that.
@asim Re: WA - that I am not sure about. I have not really tried the apps, but one thing is for sure - WA became widely known in academia (most students know about it. You can also check math.stackexchange.com - large number of questions mention WA. So, it might be an efective marketing device for M as well - time will tell)
Mathematica needs to have one strong vertical market where it is dominant. With Matlab, it is engineering. All engineering depts use Matlab much more than anything else. With Mathematica it seems to be physics and pure math. The engineering market is more huge and extends to commercial area.
@NasserM.Abbasi Mathematica as a system (at least if we look at the design and abilities) has a much brighter future than Matlab, but only provided that the market is picking what is objectively the best / most productive system. But the market has huge inertia, and the gaps in literature and deployment channels don't help things
Even though many engineering problems can be solved better in Mathematica.
@LeonidShifrin you said better than I can
You can't believe the questions I see and answer at matlab newsgroup on how to use ode45 (differential equation solver). Since it needs much more setup to use. In Mathematica, it is so easy to use.
People, I have to go. I would like to think on how to make this talk to be a real advice for Wolfram Staff. I really get sad to see MMA numbers in TIOBE or new books number. I know it's a fantastic tool, but the market don't know. There is no good model to MMA in corporate world, this should be changed.
The statistical market is mostly ceded to R, and now Julia/Numba etc. look promising. It is difficult to make a dent there, given the free alternatives.
@NasserM.Abbasi Julia, from what I could see, is very powerful, becuse it is the only existing lang. for numerical computing which does not have 2 performance scales (which happen when lots of functions have been implemented in a lower-level languages, which leads to a bunch of complex performance-tuning techniques)
@Murta I am pretty sure that WRI management has good reasons for what they do, and have some vision about the future of M. We may in this discussion be missing some important things known to them.
@LeonidShifrin, yes, it is fast since it uses LLVM compiler. But I am not talking about speed. I found the language itself confusing and not easy to grasp, say as when I first looked at Matlab which is a very easy to use and learn. With Mathematica, I am still learning it :)
@NasserM.Abbasi What I said about 2 performance scales is a fundamental thing, because it means that Julia libs and Julia itself can be fully developed in Julia only. For Mathematica, for example, this is not really true, even though large part of it is written in itself. And neither is it true for Matlab, or anything else out there.
@LeonidShifrin, good point. But I do not think speed is everything. Ease of use of a language is more import. c++ for example is very fast, but it is so complex now, now one really understand it all.
For scientific application, a language should be fast, and not too complex to use and understand. This is one reason Fortran got so popular, and also matlab.
@NasserM.Abbasi This is not really about speed. It is about the evolution. All really cool languages have a compiler written in that same language (bootstrapping) - think C, haskell, some Lisp dialects. I think this is something very fundamental for the evolution of the language.
@asim Re: Julia - it seems to be, on the surface. From what I understood, it is really a Lisp inside, made very efficient for typical numerical computing tasks.
@asim, yes, the syntax is like Matlab, but that is the top level. It has second Level, which is the complex part, the meta programming part docs.julialang.org/en/latest/manual/metaprogramming it also has many other complex features. As I said, on the surface it looks simple, but it is very complex language.
They seem to know what will appeal to ordinary users. There is work on Dataframe functionality and they have all the statistical distributions which will help for statistical stuff, and it is easy to call C.
@LeonidShifrin please look at the Julia manual here docs.julialang.org/en/latest/manual (for you, being a language expert you might not find it very complex, but for me it looks complex, I mean for Scientific applications)
@asim Yep, seems like that. I've been reading a little their dev.mailing lists, and they said to each other something like "we'd like to have it more (lisp-like), but not everyone appreciates parentheses as much as we do"
@asim No, I actually consider myself only a beginner in languages. I started coding for real may be 3-4 years ago, before that it was mostly to support my research, and no much
When I saw the personal analytics stuff from Wolfram, I thought maybe the new Mathematica will have Dataframes etc.., but unfortunately, that is not the case.
@NasserM.Abbasi I think something like Julia is the future (or anyway would like to). It's not more confusing than any other language, I think (but these things depend on the person)
@acl I agree. Because it can be developped fully in itself without loss of performance (and I mean fully,entirely), and probably can and will be bootstrapped (compiler written in itself), they've got their evolution vector all right.
@NasserM.Abbasi But not many have this property I just discussed
@acl It is interesting that when people discuss languages, they most of the time totally ignore this evolution aspect, which is crucial in the long-term. If the language can not produce a next generation of itself, its evolution becomes dependent on a bunch of external factors which are often hard to predict and control. And people will always use all available cycles and demand more.
I just finished making list of all functions in V9 with all standard packages: 3,753 functions. In version 8 I counted 2468 (But now in V9 I added Developer` and Experimental`). Will post full report in less than 30 minutes and provid link
@NasserM.Abbasi In my grad lab, my prof was the same way. despite the fact that we could code up a comparable (and understandable) solution in mma in short period of time. It did not help he was passive-aggressive about this, so my lab partner and I walked out of class one day. I still did everything in mma.
Hi! Can you tell me how to calculate following formula Ex = (P2 - P1) / ||P2 - P1|| if I have two points on plane with coordinates P1 = (39, 28) and P2 = (13, 39)
Here is the latest listing of version 9 analysis. 12000.org/my_notes/compare_mathematica/V9/main.html I list all symbols on one page, and another page contains listing of just the functions. One page contains the number of symbol per context. Hope you enjoy it. Too many functions, too little time.
79 different contexts. 7225 symbols, and 3753 functions.
Running Trace[Speak["Hello"]] and Names["*Speak*"] revealed the following possibility:
MathLink`CallFrontEnd[CurrentlySpeakingPacket]
Using this with a text that is split into a list of shorter strings allows you to interrupt the audio at well-defined points, phrase breaks, say. Here is one wa...
I think I have seen it happen, but very very rarely. Usually, it's if I'm actively checking mail or sent a mail when I'm offline. You don't need to quit and restart... just go to Mailboxes > Take all accounts online
@OleksandrR., this is the message Needs::obspct: Parallel Computing Toolkit is obsolete. Its functionality is built into Mathematica. Use Needs["Parallel`V2`"] to enable a compatibility mode with Parallel Computing Toolkit 2.1. >>
@OleksandrR. and when I do Needs["Parallel`V2`"], I get this warning: "Symbol "ParallelEvaluate" appears in multiple contexts {"Parallel`V2`", "System"};" Very confusing.
So, one message says to load "Parallel`V2`", and one does that, then another wanrning about shadow symbol with one in System context even thought I restarted Kernel also
@Szabolcs The message seems not to be online until now. Here it goes:
Concerning the Input Assistant, Ctrl+K now invokes the completion menu. But the completion menu comes up automatically when you need it, and invoking a completion is as simple as pressing Enter or Tab. I, for one, didn't really use Ctrl+K all that much before, but I am using this all the time, now. In observing other people use Mathematica, I see people getting used to using this very quickly, and it being very easy to adopt. There are a few tweaks we could stand to make, yet, but it seems to be a big win for everyone I've seen.
@NasserM.Abbasi Okay. So it seems that they have just added the message in M9, to try to dissuade people from loading the Parallel` package explicitly.
@rm-rf One, purely theoretical, situation is the following: The CodeAssist is included in the beta-release but it bugs so much, that it is impossible to use it. Then, although, a possible tester makes several reports about different issues the time is just too short, to test it very thoroughly.
@halirutan but you don't need to use it that much to see that it simply doesn't work. the only explanation I can think of is that nobody actually tried to use it
@halirutan I think the main reason is that people might've been using v9 "by the side", i.e., not for their day to day work. How often do you write a template message for a throwaway function that you just defined for an answer here?
Running Trace[Speak["Hello"]] and Names["*Speak*"] revealed the following possibility:
MathLink`CallFrontEnd[CurrentlySpeakingPacket]
Using this with a text that is split into a list of shorter strings allows you to interrupt the audio at well-defined points, phrase breaks, say. Here is one wa...
