Lightroom 2 experiments.

3D and such.

Every 10 years or so 3D movies are going to revolutionize the preferred way of ingesting cultural output -- or intellectual property as we are now told to think of it. Up until now the promise of the ultimate immersive experience has proven illusive. Beyond some enthusiasm centering around a handful of releases these fads usually die down quietly.

I can't say I expected much in terms of the 3D experience when I went to see "Avatar", and indeed, they do not appear to have cracked it this time either.

Don't get me wrong: I enjoyed the movie. I just didn't enjoy the 3D experience.

If, for a moment, we ignore the uncomfortable and clumsy Roy Orbison style glasses you have to wear, there is something more fundamental that needs to change with the way we shoot movies for this to work: cinematography.

Watching a 3D movie still feels like you are drunk. Everything is slightly blurry. Also, while shallow depth of field, motion blur and rapid movement works really well for 2D, it is awfully confusing in 3D. And what clown figured that steady-cam style shots were a good idea!?

What feels natural in 2D just makes you feel drunk in 3D. It is quite natral to move your eyes around the field of view -- perhaps more so when you are trying to give the brain the impression of actual depth. But when the director of photography has decided for you what is important and what isn't by way of shallow depth of field and selective focus, you just end up straining your eyes to focus on something you can't really focus on. I guess you have to practice to just passively accept and not strain your eyes.

(Peter Boros, who has done some work on 3D, said that a fundamental problem is that your eyes focus on something that is at a fixed distance, while in the real world, objects are at different distances along the Z-axis. This probably confuses our brain and makes the sensory experience thoroughly unpleasant. We agreed that what is needed is probably precise eye-tracking (possibly including something that can determine focal length) and real-time scene rendering using the tracking-information is probably what you need)

But it wasn't all bad. Some scenes were quite stunning in 3D. But for the most part, it just isn't anywhere near good enough yet. The movie industry needs to learn a lot more before this is going to work properly.

So what is different this time around? What is different from the previous 3D fads?

From what I gather the manufacturers of displays for TVs are all giddy with excitement about 3D these days. 3D requires better displays, higher refresh rates and a whole slew of new gear consumers would need to get in order to ...uhm..."enjoy" 3D.

It is almost like re-living the 90s when lots of really boring people were obsessed with "virtual reality". Remember that? When head-mounted displays with inertial systems were just around the corner and we would all be immersed in our own tailor made lucid dream? Well, we got a few good first-person shooters (none of which used head-mounted displays) and a whole lot of really lo-fi VR-rubbish (both with and without head-mounted displays).

Anyone else find it a bit amusing that "Avatar" uses the ultimate immersive experience as a plot device?

So will it stick? Well, if the manufacturers manage to pump out enough gadgets that are "good enough" I guess it might. At least for a while. Though I have a suspicion that the enthusiasm will die back down to sane levels once people have burnt lots of their hard-earned cash on a bunch of rubbish that doesn't really work all that well.

But hey, things happen fast these days. Perhaps we will have 3D systems that actually work well in a few years or even months?

Until then, I'll have a regular 2D movie, thank you.


Aperture vs Lightroom?

For the past 5 years I have been using Bibble Pro as my weapon of choice for dealing with RAW images from my SLR. Initially I chose Bibble for three reasons:
  • It was fast.
  • It was available on Linux, OSX and Windows
  • I liked the results
Speed is important when you shoot a lot of pictures. A typical afternoon at a race track or any sort of event results in 1000-1500 images. For a weekend I can easily end up with somewhere between 3000 and 4500 images. Most of these will be series of 2-4 captures of the same scene, so for each series of 2-4 images I usually am only interested in one capture. In my experience roughly 10% of the scenes I shoot turn out well. So per 1000 images I will typically end up with about 25-40 pictures that I want to use.

(Still, I keep the rest of the pictures because sometimes I go back and find pictures that didn't appeal to me at first, but which can be tweaked or "rescued" later).

This means that any delay of a second or more ends up costing me a lot of time in total waiting for the computer to respond. When I first go through all the pictures the app needs to be very responsive.

Bibble Pro

As mentioned before, BibblePro is fast. Very fast. Modulo some annoying batch thumbnail generation that never seems to do what I want it to do, it enables me to empty my camera and quickly start picking the pictures I want to go on working on and do some light tweaks while I am at it. Quite often I pick pictures and do 4-5 minutes of processing in the same go.

It also has some plugins I like very much -- such as the "Andy" plugin which emulates how different types of film and paper look. I have no idea how accurately it emulates various types of film, but to be honest, I don't really care because what counts is how the image ends up looking. Anyway, the "Andy" plugin itself is reason enough to pay the quite reasonable price for BibblePro and the price for the payware version of the plugin itself.

The problem with BibblePro is that they seem to be somewhat resource constrained. Currently they are working hard to get version 5 out the door (they managed to do it a few days ago) and thus version 4 doesn't have support for my compact camera (G11). While version 5 is released, cursory examination suggests that it won't be really usable for me until a lot of the things I love in version 4 are back and working. My completely unqualified guess is that BibblePro 5.0 will be usable to me sometime around Q2 2010.

With version 5 still having some way to go before it is done and Bibble 4 not supporting my G11, I have started looking at alternatives. Mind you, I have not given up on BibblePro 5.0 -- I am going to upgrade to 5.0 when it becomes usable for me if they keep the pricing as reasonable as they have so far. In fact, I would be willing to gamble that they will stay in business and pay the license fee for 5.0 now if they would let me and if it would help them.

I also find it odd that BibblePro has not gotten more press than it has. Given the atrociously bad offerings from Nikon and Canon for handling RAW images I have a hard time understanding why the press would bother talking about these pointless products and not highlight a gem such as Bibble.

Lightroom or Aperture?

Anyway the choice seems to boil down to Lightroom from Adobe or Aperture from Apple. My impression so far is that Aperture is faster and has a much cleaner, less cluttered interface than Lightroom. However, it seems that Aperture does not have a lot of forward momentum. While it does seem to tick the boxes for the sort of rapid workflow that I am accustomed to, it seems that it is somewhat light on image-twiddling features compared to Lightroom. Also, much of the apparent speed seems to be because of agressive caching and preprocessing -- which means it will use a lot more space and reviews I have read seem to indicate that the actual twiddling of image parameters does not produce the sort of immediate response I am after.

Lightroom is an Adobe product. I am not particularly fond of Adobe as a company. Some of their products are just getting more and more bloated and as a result getting less usable (such as Acrobat Reader). Other products follow a rather silly pricing scheme that does not take into account the fact that the world has changed. Photoshop is at least 4-5 times as expensive as it should be. Photo editing is not a niche market anymore. It hasn't been a niche market for some years now. Adobe need to realize this and I do have some hopes that their recent commercial failures will wake up their executives and prompt them to take another close look at the market they are in.
A digression, but an important one: Adobe being the makers of Lightroom does give me pause.

From what I can gather, Lightroom is a bit more sluggish than Aperture. It does not seem to be built for the sort of rapid browse-tweak-export type workflow I prefer. Tests done by others suggests that for a weekend worth of images I will probably spend 1-2 hours extra just waiting for the software to respond when processing a big batch of images.

On the other hand, you can do a lot more with Lightroom than you can with Aperture. The image editing features are deeper, which may mean that I won't have to resort to Photoshop to get some of the more demanding editing done. (Again, the fact that Photoshop is such a pig of a program means that I am reluctant to fire it up. If I fire it up, everything is going to grind to a halt -- even if I am on a dual CPU 3Ghz Mac Pro with 6Gb of memory. You don't fire up Photoshop to have a quick look -- you do so if you have a purpose). Having deeper editing features available in the program I use for browse-tweak-export will probably mean I will be able to do deeper editing on more pictures than what is the case now. Which might mean that I get better results and that in total I might actually end up saving some time compared to my current BibblePro workflow.


Right now I am leaning towards Lightroom. Both because it has deeper editing and because it would appear that the team working on Lightroom has more forward momentum than the Aperture team. This seems to be the opinion of many professionals and hobbyists as well since the relative market share of Lightroom is increasing. I have no idea if people are switching or if Lightroom is just growing faster. Either may be the case.

And of course, when BibblePro 5 is ready, I will upgrade my Bibble 4 license. I sincerely hope Bibble 5 will get to where it needs to be within the next couple of quarters or I fear that they will become irrelevant -- which would be a shame.


SOAP sucks

I have no idea who James Turner is (I probably should), but in his blog posting on O'Reilly Radar he mentions SOAP at the top of his list of the worst technologies of the last decade. Obviously this Turner fellow and I have something in common.

SOAP is indeed horrible. Primarily because it inherits much of the same madness that seems to be so ever-present in the XML world.

Given the fact that the XML world has had a decade to shape up, it is nothing short of shameful that XML technologies still are such a sad sight. I have yet to find a single XML related technology that does not suck in one way or another. Some of the people behind XML tools are part of the problem too. Not too long ago I wasted at least one full day on an XML library only to find that the root cause of the problems I was experiencing was the vanity of the author.

(I am not going to name him, although by willfully causing his library not to be usable within OSGi without modification, that he insists should not be done for reasons of pure vanity, he would certainly deserve being ridiculed)

And I am talking about really basic stuff here -- the sort of libraries that are supposed to make up the foundations of other technologies. How can I be enthusiastic about XML when the foundations are so rotten to the core I have to hold my nose when using them?

No kidding: I have been unable to find a single library, tool or protocol stack based on XML that is clean, clear, performant and free of quirks and unpleasant surprises. I would love nothing more than someone who actually bothers to do more than superficial tinkering to prove me wrong. Seriously, if you have it in you: prove me wrong.

But back to SOAP. As a mechanism for exposing APIs and doing remote calls, SOAP is pitiful. The interface definitions are needlessly verbose, the specifications have ambiguities, the tools are pathetic and when you do succeed in making the whole thing work, the performance is just terrible. You have to be blind and/or dumb not to want to improve on this significantly. Seriously, there is no better way of putting it and I fear that the most common reason is that most programmers aren't really serious enough about what they do to know or even care that the whole SOAP thing is pessimal.

I wish Google would open source their RPC mechanism. They have already open sourced Protocol Buffers, which are a big step forward, but which are a bit useless without the rest of the stack. To date, the stuff Google uses internally built on protobuffers is the only RPC mechanism I have seen that is practical, performant and fairly simple to use. No it isn't perfect, but it is reasonably robust and usable. SOAP, simply put, isn't.


Nokia are done.

Among people whose opinion I value, the verdict on Nokia seems to be unanimous: they are done.

Right now things look really bad for them as they flail about, unable to formulate a believable strategy. Or at least a believable strategy that they communicate externally. Here are some thoughts:

  • Ditch Symbian. Yes, I know the installed base is huge, but Symbian has no future. It is dead. Within half a decade you'll have a hard time shifting these phones in third world countries.
  • For the fastest growing mobile markets in the world the handset will also take on the role as the primary computing and internet device. This means that "smart phones" will become very cheap very quickly. Start with that premise and abort anything in the pipeline for the next 18 months that is not in line with this assumption.
  • Running with Maemo is not the best of ideas. Usually large corporations are skeptical towards fragmentation -- and then Nokia goes ahead and fragments the market. The best case scenario here is the same sort of nonsense that BluRay vs HD DVD resulted in: hesitant consumers. Grow up, contribute positively to Android and take the fact that someone from outside the mobile industry had to give you a wakeup call like grown-ups.
Remember: the iPhone has been on the market for a bit over two years. Android has been available for a bit over a year. In that time these two platforms have individually achieved more than any "innovation" done by Nokia or Sony Ericsson put together the past decade. I'm sorry, but nobody is looking to you for innovation and your current direction looks more like sheer panic than anything else. Adapt to reality or go back to manufacturing wellies.

Matter compiler

The thing that frightens me the most with the progress that is being made in biotech can best be explained by analogy.

When (programmable) computers first emerged they were mostly accessible to specialists. Chances were that if you had access to one you needed intimate knowledge of its design to make it do anything useful. Right down to how the logic was realized in hardware -- be it by way of gears, relays, valves or semiconductors. For decades computers was the domain of specialists.

As computers have become more sophisticated those who program them, on average, have become less so. In fact, I would probably not object if someone said that the software industry employs more people who have no idea what they are doing than any other industry. We spend much time bemoaning the fact that "the kids of today" are so far removed from the bare metal they can't even do simple back-of-the-envelope calculations anymore. With ever better tools and ever better programming environments any idiot can (and will) create software.

But computers seem mostly harmless.

What frightens me about biotech is when we start nearing the point where an unsophisticated person is able to synthesize arbitrary base pair sequences at a reasonable cost. Re-creating certain known pathogens would be trivial as their sequences have been published. Indeed, some pathogens are considered too dangerous to be shipped physically between labs, so they are synthesized. But worse still: far more dangerous pathogens could be designed, by accident, and under conditions that are neither controlled nor controllable. With no ill intent.

I can't really see why it should not be possible to build tools that allow unsophisticated people with no real background in biochemistry, or any adjacent field, to tinker with DNA. I bet this exact thought is what makes Bill Joy such a ... bundle of joy when he gives talks on what might lie ahead.