For the past couple of days, I have been struggling with my eyesight on a 12 inch monitor with Ubuntu Karmic. It has not helped that, for some reason or other, I have been working frequently on a Windows 7 machine. As much as I hate to admit it – Win7 beats the pants off Karmic with its eye candy and usability.

These days I am building a Linux NAS box using ZFS-FUSE on a Hardy Box and I keep running into unfixed driver regressions like this one for a Davicom Ethernet Card. I personally believe that for Lucid Lynx, all that Ubuntu should work on is getting driver support cleaned up.

However, one thing I would like to see is resolution independence (Ubuntu’s blueprint) – which is where my eyesight and the 12 inch desktop comes in. In essence what it means is that the operating system resizes it’s UI – fonts, icons, windows – at any hardware resolution to match the physical size that you are comfortable with. This does away with the problem that at higher screen resolution, fonts look smaller (because at higher resolution the pixels themselves take up lesser space). Win7 included this feature after a lot of brainstorming.

And of course, the question of high quality icons for Linux – there have been quite a few ideas submitted to Ubuntu about this – most of which rely on Scalable Vector Graphics for this. While being theoretically sound – there is a huge performance hit by using scalable graphics, rather than bitmaps for icons. OSX – which has arguably the sweetest eye candy – does not use vector graphics for its icons.

The key to this is its ICNS format – which has a set of images at varying resolutions for icons. Interestingly, GTK has had support for Apple .icns for a long time now. A very interesting discussion on the impact of resolution-independence on icons is quite lively to read.

I would be all in favor of using the Apple ICNS format as the standard supported format in Linux. Except for not-invented-here, what’s not to like ? The format itself may demand support of scaling and resolution independent technologies to be included in GTK – which I think is a good thing, rather than mutating the icon format to support the state of GTK as it is now.

One gets access to a big market of designers who create high quality icon sets for MacOS (may more out there than for Gnome) and readily available toolchains (including for Gimp)

Take a look here and tell you dont want these …. and then let’s talk.

I had picked up a 5$ copy of yesteryears’ best game – Bioshock – one which had the same bloodline as one of my all time fav games System Shock 2.

Simultaneously, I had tried out a demo of Doom3 – a game which sadly I had never bought. I have a decently respectably Core2Duo 1.8G, 2.5G RAM, nVidia Go7400 card.

Doom ran spectacularly – at 1280 X 800 – my laptop’s native resolution. I had a lot of graphics eyecandy turned on. Mouse movement was fluid. The story, of course, sucks – but Doom/Quake has never been about the story anyway. It’s the experience.

Bioshock was how 20,000 leagues under the sea would have looked, done in a 1940’s styling. The story and the plot was fabulous.

The engine, however, crawled. I used all the latest patches from Direct2Drive – but it still crawled. The Unreal engine has been no stranger to controversy regarding it’s quality – however a lot of games still use it. As I was reading wikipedia about ID Tech 4 (as Doom3’s engine is known), I found out that most game creators did not use it because of it’s early lack of expansive views (a la Farcry ). However, ID soon released the MegaTexture technology.

The ID Tech 4 engine costs about 250,000$ – I am not sure if that is a factor. However, playing both the engines side by side, showed which one was more forgiving on older hardware.

Which means Rage (with Virtual Texturing) gets my vote – although I may end up buying it for my PS3.

I wanted to examine the firmware files of a binary only third-party software. Now they had released their package as a tar.gz and makefiles assumed all libraries to be available in the system library path.

The most interesting discussion on why the developers dont want this functionality in is here. Frankly, I think the discussion is asinine – rather than discuss use cases on why I should not be given this functionality as a non-root user, why not make it simply easier. I mean even Windows does the right thing (since Vista) by incorporating Folder Virtualization.

I think there are two choices to get around the system – chroot/debootstrap method or the fakeroot method. These are quite useful to create a 32 bit toolchain on a 64 bit system as well (that is something that FatELF should take care of nicely! more about that later)

Anyway, I did neither and  quickly converted the deb into tgz to work with.

sudo apt-get install fakeroot alien

fakeroot alien -t something.deb

As my Dell XPS laptop’s video card failed (I’m looking at you – ‘olde nVidia with yer heating and packaging problems) and I had a corrupt Linux installation, I was suddenly in the mood to give Windows 7 a spin (There were a few games I still could only play in Windows).

Windows 7 truly is nice – but I was in for a rude surprise when I got ready to install Karmic Koala. I have separate home and root partitions to keep up with my ever changing base installs. As a force of habit (which comes from the Redhat 6.0 days, where you could not boot if the primary partition was not in the first 8GB), I create primary partitions for all my mounts.

When Win7 started to be installed, I had 50 GB of unutilised space lying around in my HDD. What Win7 does is create two partitions  – one for the main install and another 100mb partition for recovery. Now partition table holds a maximum of 4 entries – which means if you want more than 4 partitions, you need to have an extended partition somewhere (with anything over 4 partitions under it).

What Win7 does is silently, knock out one of the linux partitions from the partition table and creates two partitions for itself. It was catastrophic, because it knocked out my home folder. Now, I was not able to access my home partition, even by booting a linux livecd and using gparted.

After a lot of googling, I found Testdisk – a truly excellent, open source partition recovery software that works on windows and linux.

All I had to do is mark my home partition and write the partition table – this causes everything else to get screwed up (including my Win 7 partitions), but I was OK with that.

The only way around is prevention – always install windows 7 on an extended partition if you have more than one OS on it. The next time around I installed Win7, I pre-created extended partitions using gparted and installed Win7 on the extended partitions – note that the Win7 installer does not give you any sort of partition manager (the way Linux does)

The next version of Ubuntu is going to be a enterprise quality, long-term-supported version called Lucid Lynx : ver 10.04

Cant wait!

Harfbuzz is the unified text layout engine for Linux (both Qt and Pango-Gnome). Graphite is another text layout engine.

What’s the difference?

Graphite behaves somewhat like AAT – the same technology that powers Apple’s beautiful fonts in OSX. It is what is called a smart font engine. Problems occur in most East Asian languages because of the complex ways that alphabets combine to form compound alphabets (what are called maatra in Hindi). Typically, these rules are extremely complex and vary for every dialect of Hindi.

That is just one language – out of 415 in India alone.

Linux is positioned in a very unique way for the Third World countries – no foreign-exchange-bleeding drives up adoption. However, in terms of usability, it is handicapped by lack of support for most East Asian languages in a well defined way.

And this defines the attitude

For established scripts though, there is not much reason to prefer Graphite over OpenType.

English speaking countries are definitely not the turf for battle of OSes. A 1000$ Mac is more than 3 months salary (pre-tax) for a beginner software engineer in India. As a few studies put it :

From a regional perspective, Asia/Pacific is the most bullish on increasing Linux adoption, as 73 percent of respondents said they would increase deployments on the server and 70 percent on the desktop. In the Americas, 66 percent of respondents said they are either evaluating or have already decided to increase adoption of Linux on the desktop and 67 percent on the serve

I do not know if a technology like Harfbuzz can ever solve the problem of sophisticated text layout for myriad East Asian languages – but undermining Graphite because established languages do not require it, is definitely not the way ahead.

Practically, however, Harfbuzz has a lot of support gathering behind us – with Google Chrome already using it and Firefox to begin using it. What we need is integration of the Graphite rendering technology into an API compatible way inside Harfbuzz, to be enabled using a ~/fonts.conf . This prevents breakage for the English language users, while the rest of the world has a moderately usable desktop while they wait to catch up.

We need a H3W – Harfbuzz 3′rd World.

Rage, John Carmack’s latest baby – which has spawned countless papers at SIGGRAPHs and new fangled datastructures (sparse voxel octree – I’m looking at you) is almost here and the demos are looking to be amazing.

. I’ve seen no game that, in this realistic style, looks so good and has a landscape so rich with visual splendor.

Considering that PS3 is getting screwed up more and more and Carmack himself is’nt very sure about the PS3, I think people will be pushing for an upgrade for their computers. Next year is going to be pretty exciting for PC gaming with DirectX11 (on Windows 7), StarCraft 2 and Rage.

I think the netbook market is clear indication enough that the market is shifting towards thin and light and I’m interested in seeing what nVidia and ATI have in store for us.

Intel with its 32nm Arrandale chips is pointing towards a 18W TDP processor – which I am sure are going to be in the next gen macs (rather than the 4 core Clarksfield – cos they are power hungry and we cant use 4 cores efficiently yet).

More and more questions over at various notebook forums ask for a 13-14 in laptop with good gaming support. Hell, make the DVD drive optional as well (I can install Linux and Windows 7 from USB sticks). Just dont repeat the bad experience of nVidia chips overheating. A couple of my friends had that experience with their Dell XPS gaming laptops and now all of us want to go ATI – even though nVidia is so much better on Linux support.

I’m keeping my fingers crossed for next year’s upgrade window – I think Doom3 and Half-life 2 spawned sufficient upgrades. I think next year will as well.

Google announced that it would be acquiring On2 Technologies for 106 million USD. On2 owns the licenses for the VP8 video codec – the same technology which powers Flash video at its core.

Why this is interesting is because of the ongoing dispute between Apple, Google and the Mozilla foundation. Without going into the long story of the issue, the core bone of contention remains

1. Mozilla adds support for the <video> element in Firefox 3.5 (as part of the HTML5 spec) and in doing so, included the Ogg Theora codec as part of the release.
2. Google says that Theora (based on the open source On2 VP3 technology) is seriously outdated and cannot keep pace with its upgrade plans for Youtube.
3. Firefox says Theora is already better than current Youtube quality.
4. Apple says that H.264 is its preferred codec (even though its closed source and has a prohibitive license), because of availability of H.264 hardware accelerators, that would be a good fit on iPhones and other media devices.

So, as it stands today – there was pretty much a stalemate.

What I am betting is this – Google will probably open source the On2 VP8 (or more likely the VP9, VP10 technology streams) and give Mozilla a technology that is far ahead of anything else.

In one clear stroke, Google will have broken H.264 and Apple’s stronghold on technology (because other iPhone clone makers would not have to purchase expensive codec licenses to put on their devices – which only Apple could have afforded). It opens up the marketplace for hardware vendors to put out their chipsets, which will soon grow to a number to rival H.264 support.

At the end of the day, what it means for the consumer is cheap and easy access to high quality online video through any handset vendor. For Google, it means Youtube’s supremacy will remain unchallenged by a video site that Apple could have created, offering high quality video accessible through… the iPhone only !

It will be interesting to see how this battle pans out – given that Eric Schmidt no longer enjoys his board seat at Apple … for obvious reasons!

Now they have done it – when the demigod John Carmack is able to push out Rage at a higher FPS on the Xbox 360 and the lowly PC, than the PS3 (with its nVidia RSX chip), you know that they are on their gaming deathbed.

I have written about this before, and I still cant believe Sony wants to keep its higher-priced console on the same collision course. I still believe the PS3 is a great machine – however with its crippled tools and developer initiatives, it is seriously hindering its own growth.

For a long time now, PS3 has been unable to play PS2 games (and so I have never been able to play Shadow of the Colossus). Only recently, have they even got around to patenting emulating the PS2’s Emotion Engine processor on the PS3. This means they still have to get around emulating the PS2 GPU.

What makes this hard to swallow is projects like PCSX2, which fully emulates the PS2 on the PC – and all this was done by enthusiasts without access to the source code. I mean just check out their SoTC emulated gameplay.

Sony needs to get its act together, or be prepared for Betamax v 2.0.

I am not sure how practical this is, but given that verification takes place over months in chip design, this could well be interesting.

A bunch of people from the University of Michigan have created a web game, which tries to pose on-chip satisfiability (SAT) problems as a game. However, reading its description makes it sound more a DFT problem.

Essentially, each dot on a grid is controllable, which may or may not toggle a dot’s observability. Essentially, what they are doing is the exact opposite of this paper: Satisfiability-Based Automatic Test Program Generation and Design for Testability for Microprocessors. Rather than use a automatic solver for test-pattern generation, they are using humans to solve these puzzles.

I would not call it a true webgame sat solver – because that would entail a game in which you have to generate a combination where two puzzles dont match each other.

I am not entirely too sure about the objective of this study – unlike Amazon Mechanical Turk or Google Image Labeler, both of which are online initiatives which incentivizes (in the form of fun or financial compensation), small tasks which can be combined to create a meaningful whole result.

There is a technical basis for this question – unlike tagging images, test pattern generation cannot really be partitioned. If indeed it could be, then you could simply multi-thread them (or run something like MapReduce to enable distributed computing). A large enough problem, and you will have lost your ability to attract people, willing to solve the problem.

So, my conclusion is that this is an impractical, highly speculative academic exercise.