I'm not sure that the assumption that kids will be the primary users of these laptops is even valid.
I've done volunteer work in 3rd world nations and the one thing we really needed was realiable, weather-proof computers with wireless communication. The first thing I thought of when I played around with an OLPC was how great the platform was for remote areas. In these environments, the standard practice is to get hand-me-down laptops from 1st world countries. These tend to vary from barely working 386-based systems to more modern Mac and Windows machines with serious battery issues. Having a rugged, field-rechargable platform would simplify a number of applications - from simple logistics (keeping track of local resources and trade), to tracking human rights abuses (what I was doing), to the more nefarious military/rebel/drug-running uses (which is what I personally think most OLPCs will be appropriated for). I suspect adults will quickly become the primary users of XOs.
I haven't seen any serious discussion about this possible (probable) use of the OLPC and would like to see what other people think (esp. if you've worked in these areas before).
It's worth keeping in mind that the Wii and 360 are already maxed out from a development perspective (ignoring the Wiimote, of course). The (mostly) standard hardware matches the PC model, making it easy for developers to fully utilize the systems.
The PS3, on the other hand, is a radically different hardware platform that developers are only now starting to fully understand. The first generation games only scratched the surface of the PS3's potential. The second round of games will be much better, but the real fun will start once the developers have had a few rounds to learn the ins and outs of the Cell.
So, a slow start is fine for Sony as long as the developers continue to learn how to use the Cell. In 2 years, PS3 games will still look "cutting edge" while 360 games will starting showing the age of that platform.
Of course, Sony could still blow it on the online/casual games. But, at least the linux distro is giving everyone a chance to play around with the Cell.
Yeah, I wondered about this, too. Their goal is to simulate the conditions faced by Mallory and Irvine as much as possible, so removing the fixed aid makes sense. I'm assuming (hoping) that their support team will remove them and then replace them once Conrad and Leo pass. It would be a little irresponsible to not replace them and risk the lives of other climbers.
Conrad Anker is one of the best mountaineers in the world. There's no chance he's ill-perpared for this trip. Of course, they _are_ doing this trip using only 1920s-style clothes and gear (their team is removing all the fixed ropes and ladders before they ascend). But, if anyone has the skills and experience to attempt this, it's Conrad. Leo Houlding, on the other hand, is an odd choice from an high-altitude experience perspective, but he's proven himself on a number of bold ascents in Patagonia recently and this will be a great tick on his resume.
From the geek perspective, think of Conrad Anker as Richard Stallman and Leo Holding as Miguel de Icaza c. 2002.
Don't be so quick to discount this site based on its use of Flash. It's user interface is incredibly well designed and a great example of how Flash can be used to improve the user experience. It's more of an interactive media site than a traditional text-based Web site. And, amazingly, it works on my old 500 MHz G4 running Safari (very few "Web 2.0" sites work on this configuration - even most Google apps perform poorly).
Skip to the route map to get an idea of the effort they put into the site (http://www.ueverest.com/route_map.html). It's Java, not Flash (which also suggests the designers understand how to choose the right technology for the job).
Anyway, as someone who tends to have the anti-Flash bias, I was surpised by this site. Even if you're not into mountaineering, it's worth looking at to see the potential of interactive Web applications.
-Chris
(And what's up with moderators moding the anti-Flash parent as 'Insightful'?)
You're off by about 10 years.... Around the turn of the century, USB MIDI was on the scene and parallel/joystick port MIDI was widely available. Protools LE was released, Cubase and Logic were both over 10 years old and were mature on both PCs and Mac, and affordable high-resolution, multi-channel DACs (e.g. MOTU 2408) were appearing on the market. Anyone still using an Atari ST at this time was living in the past.
Whoa there... just like 'Open' has other meanings that don't imply 'Open Source', cross-platform has many interpretations. The XNA claim to be cross-platform is definitely a valid one, particularly when you consider that the X-Box 360 is a PowerPC architecture and Windows XP and Vista are primarily run on x86 chips. If single development environment that can not only target three versions of an OS, but also target multiple processor architectures isn't cross platform, then I'm not sure what is.
The best hobbyist platform is definitely the PS3. Linux runs great on it and HDMI->DVI adaptors let you use it with a DVI monitor (we run them at 1920x1600 on Dell 24" LCDs). As long as PS3s are in the $500-600 price range, there's no real incentive for another low-cost Cell platform. Of course, access to the graphics pipeline is limited, but the SPUs are much easier to program and more flexible than GPUs for general purpose computation.
If you do bite the bullet and go with a PS3, we've developed a Python library -- CorePy -- for programming the SPUs (and PPU) directly. It replaces assembly/intrinsics with Python function calls and provides components for building and optimzing SPU programs. It takes the sting out of using the C-based tools and gives you more flexibility with how you use the SPUs.
Word's interface, my point of reference for WYSIWYG, is much better than anything developed on LaTeX. Framemaker's was even better.:( (It's worth nothing that I'm not really trying to advocate Word, just pointing out that LaTeX is not the panacea that most people make it out to be)
The verbostiy of LaTeX does have a negative effect on thinking. Reading LaTeX while editing requires filtering out all the formatting codes, which consumes cycles that could be used for developing the text instead. Then there are all the whole repetitive stress issues with typing too much... but the mouse in Word is just as bad in that respect.
The LaTeX build errors show up most often in collaborative environments (which is precisely where its benefits are so tantalizing). We have a fairly large lab (about 25 people at any given time) and we share a common respository for LaTeX style files, bibliographic materials, and other Tex related components. It's easy for someone to unknowningly introduce an incompatibility in the system that goes unnoticed for months or years.
When people ask for an editable copy, it's usually because they are incorporating my work into a grant or patent application. For the former, it's usually a collaborative processes with non-CS scientists who can't handle LaTeX. If I want to benefit from the grant, I have to work a little to get my content into it. For the later, lawyer's secrataries make more than academics, so it's more cost effective for us to do the conversion (sad but true). In corporate environments, the problem never arose, since everything was done in Word (including academic papers).
Anyway, I think the point I'm really after is that neither Word or LaTeX are all they're cracked up to be. Both sides blindly defend each tool as _the_ solution, ignoring their shortcomings. Unfortunately, the status quo is good enough for both camps and, given the large code bases for each tool, the cost of addressing the core issues is probably to high at this point. Which kinda ties back into the whole Exchange killer discussion: the status quo is just fine on the Exchange side and the Open Source community has to do more than just duplicate functionality to really win over new users.
"LaTeX my friend, saying Word is a viable replacement for serious professionals is laughable"
As someone who's lived in both LaTeX and Word land, I call BS (8 years extensive LaTeX usage and currently 120 pages into a CS dissertaion in LaTeX, 8 mutually exclusive years of Word usage). LaTeX is not viable for serious professionals, no matter how often people call it a professional type-setting system.
LaTeX is about the most backwards way of producing documents there is. Compile your document before you can see the layout? What is this, 1980? Arcane, verbose commands to do simple things (text{it,bf,etc})? Might be fun for people just learning to code, but come on, the verbosity gets old fast. Build errors more difficult to debug than C++ template meta-programs? Yikes.
Yes, LaTeX does have some nice features - separate files for different parts of the document are nice as are the exensive macro expansion features - but are they really worth the hassle of dealing with the system? I use LaTeX over Word for one reason: Word can't number references, figures, and tables correctly. Unfortunately, for academic writing in CS (where they insist on [1] instead of [Smith 1994]), this is a deal breaker.
Of course, in using LaTeX, I've given up the ability to have precise control over the location of figures in my documents (trivial in Word, barely possible and not worth the effort in LaTeX). I've resigned myself to the fact that at some point during document preparation, something will go horribly wrong and I will lose half a day trying to debug LaTeX. I've accepted the fact that conference organizers will continue to give me broken style files and then complain when the formatting is a little off. I know that if anyone needs an editable copy of my document, I will have to spend a day converting it to Word. And so on...
I would happily pay for a good word processor that worked almost like Word, but got the numbering and modularity features correct (hint: Framemaker before Adobe killed it). Open Source has had at least 20 years to get this right with LaTeX (and no, none of the WYSYWIG LaTeX tools cut it) and just keeps happilly saying LaTeX is professional without ever bothering to evolve it out of the 1980s.
As much as/.'ers like to advocate spending your CS undergrad time learning math and not computing, I think its equally, if not more important, to spend your time understanding how computers work. In addition to the standard math track (calculus, discrete math, numerical analysis, probability), a well rounded CS curriculm should expose the computer science student to the inner workings of computers and software.
Do an architecture track that covers logic design and computer architecture. This will give you a solid understand of what is possible on modern hardwarend and teach you how to apply what you learned in discrete math.
Next, take an operating systems course. This will show you how software interfaces with the hardware.
Along the way, make sure to take at least two programming language courses, ideally one imperative/procedural and one functional. This will teach you there are multiple ways to solve the same problem. Once you're comfortable with languages, take a real algorithms course with a focus on analysis (not just design). If you have time, a compiler course will fill in the final gap between hardware-os-language-program.
Fill out your electives with things that interest you - graphics, HCI, databases, AI, languages, etc. Be sure to have fun in these courses and apply what you're learning in the other courses.
In the end, a good understanding of the whole stack - hardware, software, and theory - will give you the foundations to build a solid career in computer science, in industry or academia. If you want to specialize more, consider a Masters or Ph.D.
I've learned from experience in both industry and academia that a solid foundation gives you an advantge over people who specalized too early in their careers. A strong theory focus as an undergrad makes it difficult to understand what's practical, too much focus on the hardware side makes it difficult to understand what's possible, and only learning the latest trends makes it impossible to design an debug real systems. Strive for balance and breadth early, specialize later.
The point isn't that developers sell advertisements, it's that the products created by the developers bring in viewers, which in turn allows Google to sell advertisements. Its the same as TV shows: actors and directors don't sell advertisements door-to-door, either. But, their products are funded by advertisements and used to bring viewers that networks use to sell advertising time. The money to fund the developers has to come from somewhere.
Since most people use the Web to access information, the best type of products for sellling advertisements are those that "organize and parse information". But, with Web information products, there's the added advantage of being able to use very targeted advertising, adding huge value for the advertisers.
I think the YouTube purchase is informative here, particularily as a response to the altruistic notion of organizing the world's information. The content in YouTube is primarily user generated entertainment and not infromation in the grand sense of the knowledge of humankind (though it is a good cultural snapshot). But, YouTube as a Web-based application is a wonderful platform for selling advertisements to the 18-34 year old male demographic, plus a number of other demographics. If Google was purely after organizing information for the good of the world, the YouTube purchase doesn't make sense. However, if Google's goal is to sell advertisements, then it makes perfect sense. YouTube targets precisely that demographic that advertisers love and but have lost through traditional media.
Now, there is the bigger picture of how Google uses the revenues from advertising. And, in this case, I think the "organizing information" goal is probably attainable and probably is Google's meta-goal. This is indeed what Google does well. But, it's important to separate this from the business. The _business_ is selling advertisements, plain and simple.
If you really need more convincing, read Google's annual reports. They start off with the altruistic goals, but quickly dive into the business aspect of selling ads.
"Google does do only one thing, and they do it well!"
Yes. And that one thing is sell advertisements. Just like magazines. Only instead of editorials, infographics, and stories, Google uses search and nifty Web and network enabled apps to attract eyeballs.
Google's army of coders building "cute" apps are no different than a magazine's editorial staff and contirubuting writers writing targeted content that some demographic enjoys reading. Google's coders are just building content to bring viewers to the site.
Now, Google may bring all the world's information together, but that's only because it happens to help sell advertisements. If people stopped becoming interested in information, Google would look for other ways to attract viewers. Like...er...buying YouTube...
Google's business model an excellent model for developers. Using advertising to pay for developer's projects is really a holy grail for developers. You don't have to build the perfect product or meet external specs. You just need to build something cool and have fun doing it. And, you get to share it without of the traditional software strings attached. Very cool model.
Apache is the based on years of industrial development experience in network applications and modular software architectures. The original version, httpd, was nothing more than a simple implementation of a Unix daemon/server. Not a lot of innovation at that point. Apache's real innovation was not in software, but in managing to sustain an Open Source project that became an industry standard. But, without a lot of corporate charity, it would not have been sucessful.
Firefox is built on the back of the industry funded Mozilla project (AOL, Google, et al.). While technically Open Source, the enabling technologies for Firefox don't really fit the spirit of people working in their spare time to fight the man. But, as with Apache, there's not a lot of innovation in Mozilla. It's basically copying a lot of ideas that have been common in Web browsers since their inception, many of which were introduced by the commercial browsers.
IE has always been a pretty crappy application, but it took a lot of Open Source attempts to get something that could challenge it. It's dissapointing that it took as long as it did for Firefox to come along and even provide some Open Source competition for IE. It's even more disappointing that Open Source advocates insist on getting their Web browser for free - the commercial alternatives (OmniWeb, Opera (the latter being the first browser to use tabbed browsing)) are all ahead of Firefox and IE in terms of functionality and innovation.
But, that's an interesting point, which raises an interesting question about Open Source (and gives me a chance to throw out some controversial ideas for discussion:) ).
If I'm happy with the commercial solutions, and have no problem paying for them, why should I bother supporting the Open Source alternative?
It takes time to provide good feedback. The problems with Dia could be identified by simple comparative analysis between Dia/Visio/OmniGraffle. It's standard practice in industry to understand your competition, why should Open Source developers get a pass on such analyses?
I could provide that information, but it gets back to my original question: why? What's my incentive? I don't want to see the commercial offerings disappear, so I have a stronger incentive not to support the Open Source version.
I see nothing wrong with paying for software that solves a problem elegantly, which both Visio and OmniGraffle do. The development teams both provided innovative solutions that took time and effort to develop. Why not pay them for their time and effort?
Which leads to one last point: if Dia took all the good ideas from Visio and OmniGraffle and made them free, would that be a good thing? There's no innovation in copying what's already been done. But, removing the financial incentive to innovate can stifle innovation. I know that last statement is contrary to Open Source philosophy, but I'm still waiting for a truly innovative Open Source application that isn't based on years of industrial innovation.
I use OmniGraffle and Visio. OmniGraffle is very easy to use and is very good at staying out of the way when I'm building a diagram. The automated guides, mouseless editing, and tool collections are just right for drawing complex pictures. OmniGraffle is most useful for creating diagrams for presentations (though its presentation mode is a bit flaky, so I tend to export to PowerPoint) and papers. It's also useful for sketching out ideas, but starts to show its limitations as ideas get more complex and span multiple canvases. Which is where Visio comes in...
Visio lags a little on the human interaction side, but has some major functional advantages for managing complex collections of diagrams. In Visio, the diagram is simply a physical realization of an underlying model (think MVC). You can share elements across different diagrams, and have changes in one reflected in the other. You can also add custom data to any element. This is invaluable for modeling complex softwrae systems. Each element can carry its documentation with it and it's very easy to check what other components are associated with the current one thoughout the model.
Both tools are excellent. I've tried Dia and it pales in comparison - partly due to Gimpishness of the UI, but also just b/c Visio and OmniGraffle are mature applications that have the benefit of years of user feedback.
For the MS haters out there, it is worth noting that Visio was purchased by MS after it matured. Changes since the purchase have been minor.
Of course, talking Visio and OmniGraffle is way offtopic on a thread about Google... though I really hope Google doesn't do a free Visio/OmniGraffle - I would hate to see those tools replaced by a Web-based version written by an adveritsing company (ok, that was kinda on topic).
Publicly funded results should be made available, but the funding source should also provide the funding to do so. Going from the raw data used to produce the results in a research study to raw data that can be used by other researchers is an expensive process. There are a number of workflows and tools used in processing the results that will not be available to other researchers. The common FOSS argument is only use open tools, but for many scientific applications, this is not possible. Just try telling an engineer to give up Matlab for medium-scale numeric computing in favor of your favorite scripting language. They won't accept it, and for good reasons.
Instead, the funding sources (e.g. NSF, NIH, DOD, etc) should include additional support in grants for the final step of making data available in a common format. Scientists can use their favorite tools for this and commercial tools can simply support the open publication formats. Better yet, create a National Data Repository whose purpose is to handle the final data preparation and dissemation.
For publically funded software, a similar process should occur. Most research software, while useful for a very narrow set of example applications, is not developed to the point where it is usable outside these tight constraints. This is simply because there is no research incentive to go any further than "good enough for publication". Without requiring specific languages, the funding agencies should provide enough money to finish the software engineering process and enable truly reusable software results. Some labs already meet this standard, but it's not cheap (they usually have a full time development staff in addition to the grad-student and post-doc researchers).
Most scientists don't have the time or resrouces to change current process, so it's really up to the public to not only push for open data, but also suggest and support realistic approaches to the problem.
Have you looked at COM? The core of COM (Component Object Model) is a wonderfully designed binary interface for component-based development. It's clean and simple, solves the component problem in a language neutral fashion, and was the foundation for hundreds of thousands of software projects. Microsoft solved a problem that still isn't as nicely solved anywhere else. For an example of bloated component system, see CORBA.
Of course, the marketing spin, productivity tools (ActiveX, Dev Studio wizards, et. al.), and many of the component hierarchies developed with COM are bloated messes. The marketing spin (calling everything COM) led to the confusion about what COM really was. The productivity tools were just plain bad. And, most people couldn't design good component systems (which is a point of the original article).
Long double floating point (80bit)
This is just desperation. Pretty much no-one uses 80-bit floating point arithmetic IME (and yes, I do work in the field). The portability hazards and lack of true support from almost every mainstream architecture make them almost irrelevant, except perhaps for a few very small niches."
This isn't quite true. While lanugages and programmers rarely use extended precision floating point numbers directly, the registers on IA-32-based chips are 80-bit registers. All instructions operate on 80-bit values. Values that stay in registers over the course of a computation benefit from the added precision.
For most applications, this isn't terribly important. But, for the accuracy of numeric applications, it's very useful. Numeric applications that use the FPU are inherently more accuarate on IA-32 systems than other architectures.
Unfortunately, most languages don't let the programmer control when values stay in registers (the register keyword is ignored by most compilers), so any benefit from this for most software is purely accidental, or the result of hand-coded assembly.
I'm glad to see a language taking extended precision seriously. As libraries that rely on numeric code become more common (think data mining, feature tracking in images, automated car-parking algorihtms, etc), the effects of accurate computations will be visible to more end users. But, it may be too little, too late... Intel is moving away from the FPU and towards SSE, with a max of 64-bits for floating point numbers. It may be years before we get another commodity processor line that gives us a little extra accuracy for free.
For some more reading on the topic, here's a good academic rant with lots of examples and experiments that illustrate the differences between 64-bit and 80-bit floating point numbers.
"If you want exact reproduction of formatting, use PDF. Or latex."
LaTeX gives a nice illusion of exact formatting, but as a user you really don't have much control over it, unless you're willing to spend hours writing macros to force things into the proper positions. The layout algorithms in LaTeX, particularily for image placement, give the user very little control over the precise location of images. Sure, you can create complex minipage environments, but even those sometimes get placed in strange places.
Word, on the other hand, makes it very easy to precisely place images in a document and keep them there. Simply edit the image properties, set the anchors and locations and the image will stay where you want it. Takes all of 10 seconds.
Of course, Word has some real problems keeping image and reference numbers straight, so it's almost worthless for academic publishing.
The solution I want will give me the layout control that Word and real desktop publishing systems have and the modularity, programmability, and equation features of LaTeX. Sadly, this used to exist as FrameMaker...
"What do you use this for other than just to tinker with?"
High-performance (e.g. scientific, bioinformatics, financial) software development. For about $5k, you can have a teraflop cluster.
We're seeing speedups of 2-50x over VMX (AltiVec) enhanced PPC970 tools when they're rewritten for the SPUs. Of course, the challenge is rewritting apps for a new architecture. We're discovering that the memory pipelines and processors on heterogeneous multi-core processors (e.g. Cell) are different enough to warrant complete rewrites. The good news is that the performance benefit is worth the effort to refactor performance-critical portions of applications. Incidentally, we're also working on development tools to ease the refactoring process.
"...I haven't heard one good review of the system yet. They don't have a killer game.... "
For games, I completely agree. Most of the launch titles are pretty lame. However, there is one app that's made the $500 and $600 boxes worth it (yeah, we bought one of each): Yellow Dog Linux. I've been developing for the Cell BE for a few months now (on loaner systems) and have not had this much fun programming hardware since I built a processor as an undergrad. Compared to the $8-20k that the IBM and Mercury Cell systems go for, the PS3 is a bargin.
YDL on the PS3 has all the same patches and SDKs as the official FC5 Cell BE Linux distro. The PS3 has the added benefit of a graphics card. We plugged ours into the 24" Dell widescreen displays and are running at full WUXGA resolution. The graphics are framebuffered and not accelerated, but for number crunching, this is not an issue (though it would be fun to use the GPUs, too). The only downside is that only 6 SPUs are exposed from Linux (there are 8 on the high-end systems, 7 avaiable to PS3 games).
Google is not a software development firm, but an ad sales firm (check their 10-K if you have any doubts). It uses software to attract viewers in the same way television networks use programming and magazines use articles. Under this model, it makes sense to give developers a large amount of freedom to develop whatever they want. The final type/quality/status of the software doesn't matter nearly as much as the fact that there are new features appearing on the site from time to time to attract new viewers..er, users... and keep old users. Most of the applications probably won't amount to much, but just like with any media company, you only need one or two big hits a season to keep people coming back.
Google develops a large amount of its content in house in much the same way old movie studios developed all their films in house. For Google, the talent is not actors and directors but developers. Movie studios learned that you treat the talent well to keep them around and Google has taken that lesson to heart. Developers tend to want complete freedom to work on what they want with no deadlines and giving them this is the easiest way to keep them happy. Call it 'good agile development' or whatever else you want, it's really just keeping the talent happy in the hopes that they'll keep developing content to attract users.
Unfortunately, software companies that rely on software or service sales for revenue cannot take this extreme approach to agile development. They need to deliver software on occasion or someone else will replace them in the marketplace. Agile development is still the best way to go, but unbounded development only works if software isn't your primary source of revenue.
Slashdot Mirror
I'm not sure that the assumption that kids will be the primary users of these laptops is even valid.
I've done volunteer work in 3rd world nations and the one thing we really needed was realiable, weather-proof computers with wireless communication. The first thing I thought of when I played around with an OLPC was how great the platform was for remote areas. In these environments, the standard practice is to get hand-me-down laptops from 1st world countries. These tend to vary from barely working 386-based systems to more modern Mac and Windows machines with serious battery issues. Having a rugged, field-rechargable platform would simplify a number of applications - from simple logistics (keeping track of local resources and trade), to tracking human rights abuses (what I was doing), to the more nefarious military/rebel/drug-running uses (which is what I personally think most OLPCs will be appropriated for). I suspect adults will quickly become the primary users of XOs.
I haven't seen any serious discussion about this possible (probable) use of the OLPC and would like to see what other people think (esp. if you've worked in these areas before).
-Chris
Didn't Clifford Pickover's Mazes for the Mind (1994) book have a chapter on this?
(on vacation and don't have my copy handy to check...)
-Chris
It's worth keeping in mind that the Wii and 360 are already maxed out from a development perspective (ignoring the Wiimote, of course). The (mostly) standard hardware matches the PC model, making it easy for developers to fully utilize the systems.
The PS3, on the other hand, is a radically different hardware platform that developers are only now starting to fully understand. The first generation games only scratched the surface of the PS3's potential. The second round of games will be much better, but the real fun will start once the developers have had a few rounds to learn the ins and outs of the Cell.
So, a slow start is fine for Sony as long as the developers continue to learn how to use the Cell. In 2 years, PS3 games will still look "cutting edge" while 360 games will starting showing the age of that platform.
Of course, Sony could still blow it on the online/casual games. But, at least the linux distro is giving everyone a chance to play around with the Cell.
-Chris
Yeah, I wondered about this, too. Their goal is to simulate the conditions faced by Mallory and Irvine as much as possible, so removing the fixed aid makes sense. I'm assuming (hoping) that their support team will remove them and then replace them once Conrad and Leo pass. It would be a little irresponsible to not replace them and risk the lives of other climbers.
-Chris
The parent is not 'Informative'.
Conrad Anker is one of the best mountaineers in the world. There's no chance he's ill-perpared for this trip. Of course, they _are_ doing this trip using only 1920s-style clothes and gear (their team is removing all the fixed ropes and ladders before they ascend). But, if anyone has the skills and experience to attempt this, it's Conrad. Leo Houlding, on the other hand, is an odd choice from an high-altitude experience perspective, but he's proven himself on a number of bold ascents in Patagonia recently and this will be a great tick on his resume.
From the geek perspective, think of Conrad Anker as Richard Stallman and Leo Holding as Miguel de Icaza c. 2002.
-Chris
Don't be so quick to discount this site based on its use of Flash. It's user interface is incredibly well designed and a great example of how Flash can be used to improve the user experience. It's more of an interactive media site than a traditional text-based Web site. And, amazingly, it works on my old 500 MHz G4 running Safari (very few "Web 2.0" sites work on this configuration - even most Google apps perform poorly).
Skip to the route map to get an idea of the effort they put into the site (http://www.ueverest.com/route_map.html). It's Java, not Flash (which also suggests the designers understand how to choose the right technology for the job).
Anyway, as someone who tends to have the anti-Flash bias, I was surpised by this site. Even if you're not into mountaineering, it's worth looking at to see the potential of interactive Web applications.
-Chris
(And what's up with moderators moding the anti-Flash parent as 'Insightful'?)
You're off by about 10 years.... Around the turn of the century, USB MIDI was on the scene and parallel/joystick port MIDI was widely available. Protools LE was released, Cubase and Logic were both over 10 years old and were mature on both PCs and Mac, and affordable high-resolution, multi-channel DACs (e.g. MOTU 2408) were appearing on the market. Anyone still using an Atari ST at this time was living in the past.
-Chris
Whoa there... just like 'Open' has other meanings that don't imply 'Open Source', cross-platform has many interpretations. The XNA claim to be cross-platform is definitely a valid one, particularly when you consider that the X-Box 360 is a PowerPC architecture and Windows XP and Vista are primarily run on x86 chips. If single development environment that can not only target three versions of an OS, but also target multiple processor architectures isn't cross platform, then I'm not sure what is.
-Chris
The best hobbyist platform is definitely the PS3. Linux runs great on it and HDMI->DVI adaptors let you use it with a DVI monitor (we run them at 1920x1600 on Dell 24" LCDs). As long as PS3s are in the $500-600 price range, there's no real incentive for another low-cost Cell platform. Of course, access to the graphics pipeline is limited, but the SPUs are much easier to program and more flexible than GPUs for general purpose computation.
If you do bite the bullet and go with a PS3, we've developed a Python library -- CorePy -- for programming the SPUs (and PPU) directly. It replaces assembly/intrinsics with Python function calls and provides components for building and optimzing SPU programs. It takes the sting out of using the C-based tools and gives you more flexibility with how you use the SPUs.
-Chris
Word's interface, my point of reference for WYSIWYG, is much better than anything developed on LaTeX. Framemaker's was even better.
The verbostiy of LaTeX does have a negative effect on thinking. Reading LaTeX while editing requires filtering out all the formatting codes, which consumes cycles that could be used for developing the text instead. Then there are all the whole repetitive stress issues with typing too much... but the mouse in Word is just as bad in that respect.
The LaTeX build errors show up most often in collaborative environments (which is precisely where its benefits are so tantalizing). We have a fairly large lab (about 25 people at any given time) and we share a common respository for LaTeX style files, bibliographic materials, and other Tex related components. It's easy for someone to unknowningly introduce an incompatibility in the system that goes unnoticed for months or years.
When people ask for an editable copy, it's usually because they are incorporating my work into a grant or patent application. For the former, it's usually a collaborative processes with non-CS scientists who can't handle LaTeX. If I want to benefit from the grant, I have to work a little to get my content into it. For the later, lawyer's secrataries make more than academics, so it's more cost effective for us to do the conversion (sad but true). In corporate environments, the problem never arose, since everything was done in Word (including academic papers).
Anyway, I think the point I'm really after is that neither Word or LaTeX are all they're cracked up to be. Both sides blindly defend each tool as _the_ solution, ignoring their shortcomings. Unfortunately, the status quo is good enough for both camps and, given the large code bases for each tool, the cost of addressing the core issues is probably to high at this point. Which kinda ties back into the whole Exchange killer discussion: the status quo is just fine on the Exchange side and the Open Source community has to do more than just duplicate functionality to really win over new users.
-Chris
"LaTeX my friend, saying Word is a viable replacement for serious professionals is laughable"
As someone who's lived in both LaTeX and Word land, I call BS (8 years extensive LaTeX usage and currently 120 pages into a CS dissertaion in LaTeX, 8 mutually exclusive years of Word usage). LaTeX is not viable for serious professionals, no matter how often people call it a professional type-setting system.
LaTeX is about the most backwards way of producing documents there is. Compile your document before you can see the layout? What is this, 1980? Arcane, verbose commands to do simple things (text{it,bf,etc})? Might be fun for people just learning to code, but come on, the verbosity gets old fast. Build errors more difficult to debug than C++ template meta-programs? Yikes.
Yes, LaTeX does have some nice features - separate files for different parts of the document are nice as are the exensive macro expansion features - but are they really worth the hassle of dealing with the system? I use LaTeX over Word for one reason: Word can't number references, figures, and tables correctly. Unfortunately, for academic writing in CS (where they insist on [1] instead of [Smith 1994]), this is a deal breaker.
Of course, in using LaTeX, I've given up the ability to have precise control over the location of figures in my documents (trivial in Word, barely possible and not worth the effort in LaTeX). I've resigned myself to the fact that at some point during document preparation, something will go horribly wrong and I will lose half a day trying to debug LaTeX. I've accepted the fact that conference organizers will continue to give me broken style files and then complain when the formatting is a little off. I know that if anyone needs an editable copy of my document, I will have to spend a day converting it to Word. And so on...
I would happily pay for a good word processor that worked almost like Word, but got the numbering and modularity features correct (hint: Framemaker before Adobe killed it). Open Source has had at least 20 years to get this right with LaTeX (and no, none of the WYSYWIG LaTeX tools cut it) and just keeps happilly saying LaTeX is professional without ever bothering to evolve it out of the 1980s.
-Chris
As much as /.'ers like to advocate spending your CS undergrad time learning math and not computing, I think its equally, if not more important, to spend your time understanding how computers work. In addition to the standard math track (calculus, discrete math, numerical analysis, probability), a well rounded CS curriculm should expose the computer science student to the inner workings of computers and software.
Do an architecture track that covers logic design and computer architecture. This will give you a solid understand of what is possible on modern hardwarend and teach you how to apply what you learned in discrete math.
Next, take an operating systems course. This will show you how software interfaces with the hardware.
Along the way, make sure to take at least two programming language courses, ideally one imperative/procedural and one functional. This will teach you there are multiple ways to solve the same problem. Once you're comfortable with languages, take a real algorithms course with a focus on analysis (not just design). If you have time, a compiler course will fill in the final gap between hardware-os-language-program.
Fill out your electives with things that interest you - graphics, HCI, databases, AI, languages, etc. Be sure to have fun in these courses and apply what you're learning in the other courses.
In the end, a good understanding of the whole stack - hardware, software, and theory - will give you the foundations to build a solid career in computer science, in industry or academia. If you want to specialize more, consider a Masters or Ph.D.
I've learned from experience in both industry and academia that a solid foundation gives you an advantge over people who specalized too early in their careers. A strong theory focus as an undergrad makes it difficult to understand what's practical, too much focus on the hardware side makes it difficult to understand what's possible, and only learning the latest trends makes it impossible to design an debug real systems. Strive for balance and breadth early, specialize later.
-Chris
...if you're doing it from Python
-Chris
The point isn't that developers sell advertisements, it's that the products created by the developers bring in viewers, which in turn allows Google to sell advertisements. Its the same as TV shows: actors and directors don't sell advertisements door-to-door, either. But, their products are funded by advertisements and used to bring viewers that networks use to sell advertising time. The money to fund the developers has to come from somewhere.
Since most people use the Web to access information, the best type of products for sellling advertisements are those that "organize and parse information". But, with Web information products, there's the added advantage of being able to use very targeted advertising, adding huge value for the advertisers.
I think the YouTube purchase is informative here, particularily as a response to the altruistic notion of organizing the world's information. The content in YouTube is primarily user generated entertainment and not infromation in the grand sense of the knowledge of humankind (though it is a good cultural snapshot). But, YouTube as a Web-based application is a wonderful platform for selling advertisements to the 18-34 year old male demographic, plus a number of other demographics. If Google was purely after organizing information for the good of the world, the YouTube purchase doesn't make sense. However, if Google's goal is to sell advertisements, then it makes perfect sense. YouTube targets precisely that demographic that advertisers love and but have lost through traditional media.
Now, there is the bigger picture of how Google uses the revenues from advertising. And, in this case, I think the "organizing information" goal is probably attainable and probably is Google's meta-goal. This is indeed what Google does well. But, it's important to separate this from the business. The _business_ is selling advertisements, plain and simple.
If you really need more convincing, read Google's annual reports. They start off with the altruistic goals, but quickly dive into the business aspect of selling ads.
-Chris
"Google does do only one thing, and they do it well!"
Yes. And that one thing is sell advertisements. Just like magazines. Only instead of editorials, infographics, and stories, Google uses search and nifty Web and network enabled apps to attract eyeballs.
Google's army of coders building "cute" apps are no different than a magazine's editorial staff and contirubuting writers writing targeted content that some demographic enjoys reading. Google's coders are just building content to bring viewers to the site.
Now, Google may bring all the world's information together, but that's only because it happens to help sell advertisements. If people stopped becoming interested in information, Google would look for other ways to attract viewers. Like...er...buying YouTube...
Google's business model an excellent model for developers. Using advertising to pay for developer's projects is really a holy grail for developers. You don't have to build the perfect product or meet external specs. You just need to build something cool and have fun doing it. And, you get to share it without of the traditional software strings attached. Very cool model.
-Chris
Apache is the based on years of industrial development experience in network applications and modular software architectures. The original version, httpd, was nothing more than a simple implementation of a Unix daemon/server. Not a lot of innovation at that point. Apache's real innovation was not in software, but in managing to sustain an Open Source project that became an industry standard. But, without a lot of corporate charity, it would not have been sucessful.
Firefox is built on the back of the industry funded Mozilla project (AOL, Google, et al.). While technically Open Source, the enabling technologies for Firefox don't really fit the spirit of people working in their spare time to fight the man. But, as with Apache, there's not a lot of innovation in Mozilla. It's basically copying a lot of ideas that have been common in Web browsers since their inception, many of which were introduced by the commercial browsers.
IE has always been a pretty crappy application, but it took a lot of Open Source attempts to get something that could challenge it. It's dissapointing that it took as long as it did for Firefox to come along and even provide some Open Source competition for IE. It's even more disappointing that Open Source advocates insist on getting their Web browser for free - the commercial alternatives (OmniWeb, Opera (the latter being the first browser to use tabbed browsing)) are all ahead of Firefox and IE in terms of functionality and innovation.
-Chris
No.
:) ).
But, that's an interesting point, which raises an interesting question about Open Source (and gives me a chance to throw out some controversial ideas for discussion
If I'm happy with the commercial solutions, and have no problem paying for them, why should I bother supporting the Open Source alternative?
It takes time to provide good feedback. The problems with Dia could be identified by simple comparative analysis between Dia/Visio/OmniGraffle. It's standard practice in industry to understand your competition, why should Open Source developers get a pass on such analyses?
I could provide that information, but it gets back to my original question: why? What's my incentive? I don't want to see the commercial offerings disappear, so I have a stronger incentive not to support the Open Source version.
I see nothing wrong with paying for software that solves a problem elegantly, which both Visio and OmniGraffle do. The development teams both provided innovative solutions that took time and effort to develop. Why not pay them for their time and effort?
Which leads to one last point: if Dia took all the good ideas from Visio and OmniGraffle and made them free, would that be a good thing? There's no innovation in copying what's already been done. But, removing the financial incentive to innovate can stifle innovation. I know that last statement is contrary to Open Source philosophy, but I'm still waiting for a truly innovative Open Source application that isn't based on years of industrial innovation.
-Chris
I use OmniGraffle and Visio. OmniGraffle is very easy to use and is very good at staying out of the way when I'm building a diagram. The automated guides, mouseless editing, and tool collections are just right for drawing complex pictures. OmniGraffle is most useful for creating diagrams for presentations (though its presentation mode is a bit flaky, so I tend to export to PowerPoint) and papers. It's also useful for sketching out ideas, but starts to show its limitations as ideas get more complex and span multiple canvases. Which is where Visio comes in...
Visio lags a little on the human interaction side, but has some major functional advantages for managing complex collections of diagrams. In Visio, the diagram is simply a physical realization of an underlying model (think MVC). You can share elements across different diagrams, and have changes in one reflected in the other. You can also add custom data to any element. This is invaluable for modeling complex softwrae systems. Each element can carry its documentation with it and it's very easy to check what other components are associated with the current one thoughout the model.
Both tools are excellent. I've tried Dia and it pales in comparison - partly due to Gimpishness of the UI, but also just b/c Visio and OmniGraffle are mature applications that have the benefit of years of user feedback.
For the MS haters out there, it is worth noting that Visio was purchased by MS after it matured. Changes since the purchase have been minor.
Of course, talking Visio and OmniGraffle is way offtopic on a thread about Google... though I really hope Google doesn't do a free Visio/OmniGraffle - I would hate to see those tools replaced by a Web-based version written by an adveritsing company (ok, that was kinda on topic).
-Chris
Publicly funded results should be made available, but the funding source should also provide the funding to do so. Going from the raw data used to produce the results in a research study to raw data that can be used by other researchers is an expensive process. There are a number of workflows and tools used in processing the results that will not be available to other researchers. The common FOSS argument is only use open tools, but for many scientific applications, this is not possible. Just try telling an engineer to give up Matlab for medium-scale numeric computing in favor of your favorite scripting language. They won't accept it, and for good reasons.
Instead, the funding sources (e.g. NSF, NIH, DOD, etc) should include additional support in grants for the final step of making data available in a common format. Scientists can use their favorite tools for this and commercial tools can simply support the open publication formats. Better yet, create a National Data Repository whose purpose is to handle the final data preparation and dissemation.
For publically funded software, a similar process should occur. Most research software, while useful for a very narrow set of example applications, is not developed to the point where it is usable outside these tight constraints. This is simply because there is no research incentive to go any further than "good enough for publication". Without requiring specific languages, the funding agencies should provide enough money to finish the software engineering process and enable truly reusable software results. Some labs already meet this standard, but it's not cheap (they usually have a full time development staff in addition to the grad-student and post-doc researchers).
Most scientists don't have the time or resrouces to change current process, so it's really up to the public to not only push for open data, but also suggest and support realistic approaches to the problem.
-Chris
"BLOAT - Look at J2EE and COM"
Have you looked at COM? The core of COM (Component Object Model) is a wonderfully designed binary interface for component-based development. It's clean and simple, solves the component problem in a language neutral fashion, and was the foundation for hundreds of thousands of software projects. Microsoft solved a problem that still isn't as nicely solved anywhere else. For an example of bloated component system, see CORBA.
Of course, the marketing spin, productivity tools (ActiveX, Dev Studio wizards, et. al.), and many of the component hierarchies developed with COM are bloated messes. The marketing spin (calling everything COM) led to the confusion about what COM really was. The productivity tools were just plain bad. And, most people couldn't design good component systems (which is a point of the original article).
-Chris
This is just desperation. Pretty much no-one uses 80-bit floating point arithmetic IME (and yes, I do work in the field). The portability hazards and lack of true support from almost every mainstream architecture make them almost irrelevant, except perhaps for a few very small niches."
This isn't quite true. While lanugages and programmers rarely use extended precision floating point numbers directly, the registers on IA-32-based chips are 80-bit registers. All instructions operate on 80-bit values. Values that stay in registers over the course of a computation benefit from the added precision.
For most applications, this isn't terribly important. But, for the accuracy of numeric applications, it's very useful. Numeric applications that use the FPU are inherently more accuarate on IA-32 systems than other architectures. Unfortunately, most languages don't let the programmer control when values stay in registers (the register keyword is ignored by most compilers), so any benefit from this for most software is purely accidental, or the result of hand-coded assembly.
I'm glad to see a language taking extended precision seriously. As libraries that rely on numeric code become more common (think data mining, feature tracking in images, automated car-parking algorihtms, etc), the effects of accurate computations will be visible to more end users. But, it may be too little, too late... Intel is moving away from the FPU and towards SSE, with a max of 64-bits for floating point numbers. It may be years before we get another commodity processor line that gives us a little extra accuracy for free.
For some more reading on the topic, here's a good academic rant with lots of examples and experiments that illustrate the differences between 64-bit and 80-bit floating point numbers.
-Chris
"If you want exact reproduction of formatting, use PDF. Or latex."
LaTeX gives a nice illusion of exact formatting, but as a user you really don't have much control over it, unless you're willing to spend hours writing macros to force things into the proper positions. The layout algorithms in LaTeX, particularily for image placement, give the user very little control over the precise location of images. Sure, you can create complex minipage environments, but even those sometimes get placed in strange places.
Word, on the other hand, makes it very easy to precisely place images in a document and keep them there. Simply edit the image properties, set the anchors and locations and the image will stay where you want it. Takes all of 10 seconds.
Of course, Word has some real problems keeping image and reference numbers straight, so it's almost worthless for academic publishing.
The solution I want will give me the layout control that Word and real desktop publishing systems have and the modularity, programmability, and equation features of LaTeX. Sadly, this used to exist as FrameMaker...
-Chris
"What do you use this for other than just to tinker with?"
High-performance (e.g. scientific, bioinformatics, financial) software development. For about $5k, you can have a teraflop cluster.
We're seeing speedups of 2-50x over VMX (AltiVec) enhanced PPC970 tools when they're rewritten for the SPUs. Of course, the challenge is rewritting apps for a new architecture. We're discovering that the memory pipelines and processors on heterogeneous multi-core processors (e.g. Cell) are different enough to warrant complete rewrites. The good news is that the performance benefit is worth the effort to refactor performance-critical portions of applications. Incidentally, we're also working on development tools to ease the refactoring process.
-Chris
"...I haven't heard one good review of the system yet. They don't have a killer game.... "
For games, I completely agree. Most of the launch titles are pretty lame. However, there is one app that's made the $500 and $600 boxes worth it (yeah, we bought one of each): Yellow Dog Linux. I've been developing for the Cell BE for a few months now (on loaner systems) and have not had this much fun programming hardware since I built a processor as an undergrad. Compared to the $8-20k that the IBM and Mercury Cell systems go for, the PS3 is a bargin.
YDL on the PS3 has all the same patches and SDKs as the official FC5 Cell BE Linux distro. The PS3 has the added benefit of a graphics card. We plugged ours into the 24" Dell widescreen displays and are running at full WUXGA resolution. The graphics are framebuffered and not accelerated, but for number crunching, this is not an issue (though it would be fun to use the GPUs, too). The only downside is that only 6 SPUs are exposed from Linux (there are 8 on the high-end systems, 7 avaiable to PS3 games).
-Chris
Google is not a software development firm, but an ad sales firm (check their 10-K if you have any doubts). It uses software to attract viewers in the same way television networks use programming and magazines use articles. Under this model, it makes sense to give developers a large amount of freedom to develop whatever they want. The final type/quality/status of the software doesn't matter nearly as much as the fact that there are new features appearing on the site from time to time to attract new viewers..er, users... and keep old users. Most of the applications probably won't amount to much, but just like with any media company, you only need one or two big hits a season to keep people coming back.
Google develops a large amount of its content in house in much the same way old movie studios developed all their films in house. For Google, the talent is not actors and directors but developers. Movie studios learned that you treat the talent well to keep them around and Google has taken that lesson to heart. Developers tend to want complete freedom to work on what they want with no deadlines and giving them this is the easiest way to keep them happy. Call it 'good agile development' or whatever else you want, it's really just keeping the talent happy in the hopes that they'll keep developing content to attract users.
Unfortunately, software companies that rely on software or service sales for revenue cannot take this extreme approach to agile development. They need to deliver software on occasion or someone else will replace them in the marketplace. Agile development is still the best way to go, but unbounded development only works if software isn't your primary source of revenue.
-Chris