Almost nobody does real 'Beowulf' style clusters anymore, if you use a very precise definition, but in the broader sense this would be a compute cluster. I haven't studied the details of the proposal, but I'm not sure why this would ne revolutionary. Using a cluster for rendering video is pretty common. Burn clusters are used with Autodesk Smoke. Smoke used to be quite obscure but now that it exists on Mac it is seeing pretty wide use. From what I understand, Vegas has network rendering as well. Outside of strict 'editing' and into video comp and color, tons of people use things like Nuke and After Effects on a farm and Baselight uses a cluster. So, the idea isn't really revolutionary. It all boils down to whether or not the implementation is good or not, how it gets used, and wether or not it actually makes my life better.

Distributed processing for video editing is quite tricky. Unless you are doing a lot of fancy effects (which tends to happen in something like Smoke, but as far as I know Openshot has less in the way of professional finishing effects. Some of this would include cheesy stuff like lens flares. Most o it would be stuff like tracked stabilisation and degraining which can be quite slow) actually coming up with a video frame isn't that CPU intensive. When doing ordinary cuts only editing, you just have to seek to the frame in a video file, and decide the frame. For a proper editing format like ProRes, this is about as CPU intensive as decoding a JPEG. For interactive editing, that's pretty much it. Schlepping video frames across the network for that is a huge waste. If the remote system doesn't have the right codec installed, you are sunk. As your timeline gets more CPU intensive, you get more of a payoff for having Extra CPUs to throw at the problem over a relatively high latency and low bandwidth link. Figuring out exactly what to farm out and when is a nontrivial task and it won't be possible to come up with a system that works well for all possible use cases.

Anyhow, I wish them luck. Hopefully they come up with a feature set that matches their users needs. I may have to check out a current version. Last time I played with it, OpenShot wasn't really my cup of tea, but it is always good to see somebody getting support from the community to scratch an itch.

On the subject of working with a big network, and the sort of creative influence that levels of executives can have on a project... Can you please start this interview with a fist fight?

(A probably too obscure reference to the start of the B5 spinoff show, Crusade.)

Or, to put it a bit more seriously, given that you have had frustrations with the role that networks have played on some of your previous projects, how do you see the creative process changing as time moves forward and technology changes the nature of distribution? What do you see as an end-game for these trends moving forward?

To be fair, if somebody tried to sell something as locked down as the iPad is during the period when IBM first released the PS/2, it would have also flopped. The market has changed a lot since the 1980's. People who seriously upgrade their desktop are a rather small fraction of the total market for programmable things with CPU's.

Unfortunately, you still need two separate deep images. One for each eye. Things like specular highlights and reflections shift around depending on the viewer's position, so generating a second view froom the deep data doesn't actually look right. There are some cases where it'll work fine, but it's not universally true.

Does an n-core, n-processor, or n-whatever computer running for 1 hour count as 1 compute-hour or n? Or some other number altogether?

Given the public relationsy nature of the number, assume they are going with whatever sounds biggest to the marketing guys, which would probably mean core-hours rather than machine-hours. That is to say, a two socket machine with 4 cores per socket can do 8 core-hours of calculations per hour of wall-clock time.

I can't be certain that this is the technique Dreamworks used, but it makes sense and would save disk space:

http://en.wikipedia.org/wiki/Deep_image_compositing

No idea how much this was used on this particular film, but for the record, deep images don't save disk space. They churn through it like nobody's business! The idea is driven around storing many samples per pixel instead of just one, so you have a *lot* more data than with a normal "shallow" render and compositing pipeline. It is extremely useful for some thing, but it costs so much in terms of resources that it is still not in wide use yet. On the film Battleship, ILM used a deep image pipeline, but they only did it on a handful of shots where it was deemed worth it. The vast majority of the film was done with normal issues since trying to go 100% deep images would have simply crushed ILM.

Things like 3D assets, textures, etc. don't suddenly need to be duplicated. In fact, the 3D scene itself needs very little changes, just having two cameras instead of one. It's once the movie's rendered that things double in size, but that's only a subset of the total movie's required space.

While it's true that most 3D scenes don't need major changes, and the camera data is very small, this isn't always true. In 2D, artists will occasionally use horribly nonphysical hacks to make something look the way they want,a nd some of this can break horribly in stereo. So, every once in a while a cheap flat card with a texture on it gets replaced with a real fleshed out 3D model, or a 2D dust/sparkles particle effect gets redone as a full 3D simulation. Stuff like muzzle flashes, glows, halos, flares can be quite easy in 2D but surprisingly more than 2x the work to do in stereo without giving you a headache.

You would think that for stereoscopic imagery instead of single-viewpoint imagery that the data-storage requirements would double rather than increasing by 30%. Maybe there's compression of imagery involved to save that space?

Generally speaking, nothing clever happens for stereo storage. It's just that the actual rendered frames are only a small part of the total data involved in making a film. I've never worked at Dreamworks so I can't speak in detail about their pipeline, but I wouldn't be surprised if they are rendering to .sxr or something similar. (Possibly an analagous in house format.) It's basically an OpenEXR file with metadata conventions for stereo. Basically, the two views in SXR are completely separate images stuffed into a single file. No correlation between the views so that pixels that are visible from both can be reused. You can read the SXR spec to learn all the gory details.

Microsoft. Basically, when Elop took over, Nokia became an MS Vassal. That's when they dumped the world's most popular phone OS and their internal modern OS development projects for Windows Phone, and why Windows Phone ads use Nokia phones. It's basically the same play they ran when they got SGI to start building NT workstations. And, not that far off from the investments in SCO to enable the fight against Linux. Note that the MS Vassal is actively using their patent portfolio specifically to fight one of Google's strategic plays, despite the fact that a phone vendor that has given up on OS development would probably do much better if they added Android to their phone portfolio.

Just for the record, no creative professionals use dedicated and expensive tools to visualize color spaces. If they use an expensive tool like Maya for it, it's because they happen to have it handy for more sensible purposes. Visualizing color spaces is really just a novelty for most people. Anybody who needs to do it regularly isn't so much a "creative" professional, as a color scientist.

Still, sort of a neat demo of the Blender Python API.

Is the device "smart"? I was imagining that a device like this would contain accelerometers to help control the flow of plastic out the tip - the faster/slower you move the tip through the air, the faster/slower the plastic is spit out. That would make the flow more "natural" and reduce the risk of the plastic globbing up or being stretched too thin, which is what happens with glue guns.

Nope, not smart in that sense. As far as I know, the design has no accelerometers. It's a super neat thing, but you basically have to be quite deliberate in the speed that you draw things with it. It's unlikely that what you suggest would work as well as you'd like because changing the feed rate would change the time spent on the hot element being melted, so moving slowly would be harder to cool effectively, and moving fast would mean semi-melted plastic. You can obviously change the wattage applied to the heating element, but actually changing the heat in something with enough mass to be useful for the application, quickly enough to work the way that you want, would be very hard.

Surely this goes totally against the main advantage of 3D printing - create a complex shape in CAD and click print - no crafting knowledge/skill necessary! You get accuracy and get to go do other stuff while your creation is being printed.

Obviously, if you want to make a dozen matching wallplates for light switches, you would want to use a "real" 3D printer instead of something like this. But, if you just want to doodle something for fun, then that main advantage of 3D printing doesn't really apply. Just making copies of stuff that other people have designed isn't really the only possible application of the technology behind 3D printers. Besides, if simplicity is the main selling point, not needing a computer, CAD software, fiddly motors to keep calibrated, firmware updates, or any of the other complications of a full 3D printer, is pretty neat. At very least, you have to admit that a lot of people are clearly interesting in it, considering how far they have blown past the original Kickstarter goal. (And, if you look at how small the original Kickstarter goal was compared to where they are at, you can see that they really weren't expecting quite this much attention!)

Disclaimer - I haven't actually had a chance to play with one of these things, but I am friends with some of the people involved in launching the kickstarter, so I have been following it for a while.

Well they're selling them for $50, that's not bad. And a hot glue gun doesn't have the same level of finesse/control, from what I can see online..

I haven't actually gotten to play with one personally, but I am friends with the guys involved in this. Basically, yes, there is an analogy to be made with a hot glue gun. But, have you ever tried to "draw" a cube with a hot glue gun? Good luck with that. With the 3Doodler, you are working with a much finer 3D printing style plastic filament. It's essentially the printing head of a 3D printer that you can use to draw freehand with. The plastic coming out of the head cools very quickly, and is relatively strong compared to something like glue, so you can make all sorts of interesting shapes. They may post some additional videos to help clarify how easy it is to draw with.

I read this as "Do the work for HR/Recruiters, they're not intelligent enough to do it themselves".

That's a bit unfair. There are occasional HR people who are quite intelligent. (They may be rare, but it's not an ansolute given that all HR are stupid.)

But, even the intelligent ones aren't going to do a bunch of extra work for no reason. Yes, you should always try to make it as easy as possible to hire you. HR people may have to hire candidates for 20 wildly different specialties by Thursday. They may have to hire an accountant, a DBA, a sales person, and a nurse all at the same time. If an HR person had a deep interest or understanding about IT, they would probably just be an IT person instead of HR. Somebody hands them a big list of topic-specific jargon. They try to get some sense of what sort of person they need, but the distinctions between writing SQL, Java, and ASM are almost certainly going to be mostly lost on the HR person.

Then, once the HR person gets some understanding of who they need to find, cheat sheet of keywords in hand, they get about 1500 resumes dumped on their desk. Half of them are from short order cooks who heard that "Computers" is a good business to get into. Half the remainder are from college grads and tech support weenies who have decided to fluff up their resume far beyond their abilities with huge, blatant lies. Then they still have a pile of things to sort through of people who probably could do the job. They may simply not have enough time to do more than give a cursory skimming of most of the resumes. So, if a resume has all the words they've been told to look for, it'll tend to get quickly to the top of the pile. If a resume has none of them, it'll tend to wind up at the bottom.

In bad cases, that's because the HR drone can't line up "Oracle and MS SQL Server Admin" on a resume with "Database Manager" on a request. It seems dumb, but to a person not in the specific field, the jargon can seem very opaque. So, you have to try to make it as easy as possible for somebody to hire you. I almost always tweak my resume to only mention relevant experiences whenever I apply for a job I'm actually interested in. It means I have a better signal:noise ratio for that position, and I'm more likely to sounds like what they've been told to find. It's a little more work. But, boo hoo. They probably want somebody who thinks working hard to make other people's lives easier is a good thing.

Meh. I want to do less things to get where I'm going, not more. If I have to fiddle with docking my phoen every time I get in my car, that's really not a useability win. Plus, I have a bunch of strictly automotive apps installed on my phone that I don't need when I am away from the car. I think there is a good place for an open car computer system, running car specific apps.

For example, I think it would be a good idea to keep track of my mileage and how much I spend on gas. An in-car app can detect whenever I have put gas in the tank, and pop up a little dialog asking how much I just spent. I think that app should be able to sync with my phone or whatever, but It isn't something I would sensibly have any use for on my phone when I am away from the car.

Of course, I also think we should have autostereoscopic screens in the dash, and stereo backup cameras in the rear. So, my tolerance for built in gadgetry may be higher than others'.