Fabrice Bellard is a genius. Seriously, this guy may just be one of the, if not the best software developers around. The very fact Bellard seems to like tackling heavily technical challenges and sharing his findings is why he's the great developer he is. This is impressive.
How many people do you know can do all this. There are a few threads being posted these days about side projects/extra income projects/failed ideas. Nearly 99% of all ideas were designing HTML pages. Simple MVC apps.
How many things related to hard core development do you see these days?
Those people who do real hack work, pretty much keep silent and to themselves. Don't harp, tweet or blog every two minutes.
While this stuff is very interesting, websites and software can also be quite interesting feats. Although I would agree many that you see here really aren't.
I like the model Fabrice uses. He picks really meaty, technical projects which most individuals would not attempt alone, if they even think the project is possible. And he comes up with the goods. It would be interesting to know if this model works... umm financially. Very tempting to give it a try if it does.
Getting paid is a proxy for doing something society actually wants (rather than paying lip service to) and brings the resources to make it self-sustaining. I'm torn because I find society's wants (at least around my calling) to be largely fickle and shallow, but if I disregard them it's mere self-gratification.
I'm pretty sure that the market would support televised gladiatorial combat, but then you remember ethics. And if you'll allow ethics, why not also allow other considerations than purely what will make you money? Rachmaninov could have just written vaudeville tunes and radio advert jingles for lots more money, but we'd have not have heard of him today if he had. Was he merely self-gratifying, or are there more dimensions, other than what will make you richest precisely right now, in this 'worthiness' space within which you're maxima-finding?
Unfortunately, one has to eat. And there are social responsibilities, e.g. to family and the needy and so on. It's tempting to work on projects that are really meaningful, but if one receives nothing at all for it, the result may be starvation or a life of crime, unfortunately.
It's sort of a proxy for what society wants, but a rather noisy and inaccurate proxy. For one, it's only a proxy for the proportion of value you can personally capture: value you produce for society but cannot capture a good percentage of is still providing things that society wants, but not increasing your bank account. And for two, it's only really a useful proxy of short-term value: if you're doing work that society is going to love in 20 years, that doesn't help you eat today (which is why basic research rarely pays as a commercial business model, because the value it creates isn't near-term enough).
It's all to easy to convince oneself into creating things society should or may eventually value. In my case it took a long time for reality and the absence of evidence to sink in.
I think it's also far too easy to convince oneself into believing that a profitable business is actually creating value, rather than merely moving it around. So there are many pitfalls.
What about all the stuff you have to work through that society doesn't want to pay you for* before you can create something that they do? Life isn't a binary "Will I get paid or will I indulge myself with this project?".
*I am extremely loathe to use no payment as a direct indicator of want or value.
I was pleasantly surprised to see the LTE specs are freely available. Large consortium specs of this nature seem to be quite expensive. (cf: http://www.iso.org/iso/home/store.htm)
Wonder what the patent licence situation is with the LTE specs?
Software that has no 'technical effect' (though details of the specific test have changed a few times) are not allowed. Many many valid [in respect of patentability of the subject invention at least] software patents have been grantd by the EPO and EPC member nations.
Because business methods aren't patentable in Europe many USPTO granted patents have no technical contribution that is not excluded from patentability.
tl;dr, it's complicated but your statement is false.
Interesting concept. So if someone has a "hardware" patent and you manage to implement a software "emulation" of that hardware (that nevertheless runs at full speed) then you don't have to pay?
Look at his ouerve, he doesn't bother integrating any of his projects into anything else, or put more charitably, dance to the politics of other peoples' projects.
I don't blame him. Contributing to GNU projects as an outsider is a PITA.
It's more than that, I think: I've worked with people of his caliber in other fields, and I can appreciate that they need to work at their own pace.
e.g., if he submitted QEMU as a patch for bochs, and it was rejected (for whatever reason), does anyone think his time was better spent fixing that patch?
He produces ingenious, high quality code, and does so at an unbelievable rate (note, none of his open source projects is related to his day job as far as I can tell - except perhaps the ASN 1 compiler).
I think he is arguably optimizing his contribution to society, and I hope he keeps doing that. Let people who can't produce the core do all the integration/politics/detail stuff.
What does he do for a living? I can't find anything about it on the wikipedia page, his personal page and the blog page about his achievements - strange.
Cool, it's from the same author (Fabrice Bellard) who coded the PC emulator in Javascript: http://bellard.org/jslinux/
("How much time takes your browser to boot Linux?")
Very neat. Shame that 'interested companies can contact me'.
Would be fun to know how such base-stations - when attached to antenna - can be used to spoof things and perform attacks. Lets hope that comes up at some blackhat conference soon...
I would not be surprised if his intent is to first make money on his insanely hard and smart work and then make it open source later on. I think he has done something similar with another project of his - I just don't recall which one.
Yeah, kqemu. As far as the reason, that definitely might be it. KVM killed the need for kqemu. May be Harald Welte and co will release a GPL implementation which will force Ballard to open source his. Evil genius :)
Neat, however, software for base station are not provided and interested companies can contact the site owner. Pretty cool that this is allegedly running on a software defined radio kit which can be had for only ~$2000 (a bargain).
Because you are comparing two different categories of radios. There is a big difference between a very cheap receiver and a high quality, wide bandwidth, customizable transceiver. A bad automobile analogy would be a Yugo vs Peterbilt.
Wonder how RealTime it is? Encoding/Decoding of data/signal as per LTE specs is a multi-step process with lots of maths intensive operations.
Normally, custom made Baseband DSP processors are used for this kind of stuff which contain special HW Accelerators for this kind of intensive computing.
Would be really interesting to see how he has implemented it and how practical is it.
Had it been somebody else, I would have pondered on the truth of the claim. But Mr. Bellard... that is another thing entirely.
My guess is that he does have certain limitations, for instance the amount of connections he can manage and so on. But I do think modern hardware can be made to perform well if you know what you are doing. There are many ways to generally implement the DSPs on a modern PC that would be fast enough.
Also, a guess is that one of the reasons DSPs are preferred is that they have a better power profile. You don't need that for testing purposes.
It uses a USRP N210 (https://www.ettus.com/product/details/UN210-KIT) which has an FPGA that is most likely used for the signal processing. However, quite a lot can be done on a standard PC these days with SSE or GPGPU programming.
I have a question here.. No one doubts the genius of fabrice bellard.. Someone in the thread rightly commented him as a polymath.. Fabrice sticks to C and loves the language to the core.. I often hear the FP folks saying that ppl who are inclined towards math love to program in lisp/scheme/haskell/clojure.. Fabrice Bellard being a math genius has no project written in any of these languages.. Why so? I understand it is personal taste but there got to be deep reason also
Several of his projects show he has an affection for applied number crunching and low-level programming. C lives right at the intersection of these domains.
Productivity and sheer balls. My on-the-side projects are pathetic compared to what this guy does. Even a quick glance at one of those LTE specs is enough to scare me a million miles away from a project like this.
Knowing how many people it takes to develop a eNodeB is even more impressive. Normally it would be n * 100 engineers (or even n * 1000) working on these projects.
Considering that this is from Fabrice Bellard, I actually was quite disappointed that this isn't really software-only, i.e. uses the side effect radio waves generated by some standard hardware to do it. Talk about high expectations…
You can't generate an electromagnetic field just from software!! At some point you need a physical medium to accomplish that. In the case you point out he uses the DAC hardware on a graphics card to generate electric signals, controlled by software. Not really sure what you meant.
That's like saying "you can't calculate pi just from software". We're willing to presume the presence of some type of general-purpose hardware when we talk about software.
Some of his projects require only the use of commodity hardware of the type sold at Walmart. That's categorically different than something that requires a custom built software defined radio transceiver.
While it's a big technical achievement, in a world when companies are talking about 4G LTE bill of materials of $50 [1] for a picostation(roughly 20mhz) and vodaphone setting targets for $100(my guess for a full device) , where does this fit in ?
I don't think it's the main reason. If he really wanted to open source it he could do it anyway. What can the respective patent holders do? "Bring it on", get bad press, no money (assuming he doesn't make any off this) and Streisand effect.
Assuming he does want to make money off this, it makes sense to hold source (or indeed the whole prototype) closed for many reasons.
As cool as this is, the title is a bit misleading. This setup requires a "low cost software radio frontend" costing about $2000 and the author is using a high end i7 processor.
Still a very cool use of SDR and probably considerably cheaper than the alternatives.
$2k is exceedingly cheap in the cellular space I would guess. So, the title would only be misleading out of context, in which Fabrice assumes you will find your own context, because he is to busy saving the world with code.
Exactly. Go on ebay and search for "3g base station" or "cdma test" and so on ... low to mid 5-digit price tags for old equipment. Even the iden bases (nextel, etc.) are still fairly expensive...
As the documentation says, all the processing from physical to protocol layers is done by the PC. Could anyone experienced in wireless technologies explain what is the software radio frontend worth 2000$ actually doing?
A USRP costs almost $2,000 because it is completely general-purpose, modular, and customizable. It's more flexible than you really need, but if you're doing a one-off project it's cheaper than designing a custom radio.
so right now the only use of it is to get the raw packets from ethernet, convert to analog with required power and just release it over the metal? I may be wrong here.. not an RF expert
Cool! The wall around another walled garden starts to crumble! How many years before the carriers themselves start using open source code in their infrastructure because it's more reliable and robust?
This particular garden has legal walls: 4G radio frequencies are licensed. Transmitting on them without a license can (depending on circumstances) result in jail time.
I thought the frequencies necessary for LTE in the US are licensed exclusively to carriers? Or does LTE tolerate different networks on the same frequency?
They are, which makes the carriers "primary users" of those bands. As an unlicensed user, you may still legally transmit on these bands as long as your transmissions are limited to a small range and do not interfere with a primary user's use.
This means that you must transmit at a low power, but exactly how low depends on the band.
Does he need money? One would think he probably has a pretty good day job already. But if he really needed it in order to continue writing software, I would bet many people would be willing to contribute to a "Bellard Writing Fund". His contributions to open source are really in a class by themselves.
Individual persons can't deploy their own LTE base stations, unless they've got a couple million in their pocket for the radio frequency licenses. The only way this will be available to the consumer is if a telco produces it.
I'm clueless about the underlying legalities/politics, so was wondering: could something like this be used to provide connectivity in an area without 4G (but with an Internet connection)? In other words, could I (an average Joe) set up my own "cellular hot spot" and offer a better quality signal to people in the neighborhood?
No doubt he is a genius, but what has made him excel is discipline. Try to look at the documentation of the tools that he created for this project, you will understand what I mean :)
