aporem.net

In theory, programmers always knew that creating monolithical software creates problems on the long run. But the short time gains were so alluring and sometimes one got lucky and the increases in hardware speed concealed the bad design decisions.

But now this is no more. Switching to a distributed system architecture has become inevitable once the amount of data or computation requirements reach a certain limit.

For this, programmers need to accept that they have painted themselves into a corner.

I’m still very skeptical that KDE 4 will be released as scheduled. But apparently the developers have not curbed their ambitions and are planning to release 4.0.0 by December 20.
My bet is that if they keep the timeline KDE 4.0.0 will not be a complete product.

I do not know why one would need a personal portable super computer, but sure it is a cool thing: the Microwulf 26Gflops cluster build for only USD 1,256.

A small step for the implementors, a huge step for the users: BulletProofX!

It’s simple as cake: if for some reason your Xorg fails to start up, instead of sending you to the text console the BulletProofX will start a VESA X11 and reduced capabilities to give non-consolists a chance.

I could not agree more. Full feeds are important.

The argument for full feeds is that it allows a reader to be more efficient because they can digest more information per unit time. At least that is true for me. The other big benefit is that it allows people who want / need to read offline to do so. The question is, “Doesn’t reader efficiency come at the expense of the publisher”? My answer is, no, not if your content is good. In fact, if your content is good, reader efficiency works in your favor. If your content is good, then you as the publisher doen’t want me to have to break my workflow (by switching to a browser, browser tab, or NetNewsWire tab) to determine that the content is good. If I have to break the flow, there’s much less chance that I will do command-shift-P (in NetNewsWire) to pop the your post into Ecto where I can quote it as part of my post (which ought to generate some additional traffic for you). There’s less chance that I will hit command-1 to pop your post’s title and permalink into Twitterific, where it can get pumped into the realtime information junkie network. And there’s less chance that I will hit command-control-’ to pop your post’s permalink into Pukka where I can quickly tag it and stick it into del.icio.us, where it can be immortalized as important, seen by my del.icio.us network, and pumped into my blog and tumblog. In other words, you make it hard for people like me to help you. Now you might not care about that, and that’s a completely rational choice. But since just about everything in the blogosphere (after your good original content) is about getting flow (which doesn’t just mean inbound clicks) from other people, it seems like a short sighted thing to make it hard for flow to happen.

Full article here.

Here is a list of almost dead but once popular programming languages. — Something different than the usual “currently most popular (read: hyped) languages” rankings.

Programming a multi-threaded application becomes more and more a common challenge since the dawn of multi-core processors. The problem programmers have to confront in this area are manyfold. First, one can not longer automatically benefit from Moore’s Law (i.e. the doubling of transistors on a chip every 12 to 18 months) since more cores mean more power but only if every core can be put to work. Second, the redesign needed to make use of more than one core often raises the issue of a complete rewrite — with all its dangers and implications. And there are the problems arising from the lack of tools and libraries. How does one implement locking of shared ressources? What about all the variables?
This is where Transacional Memory comes in. It tries to solve the problem of multiple threads running in a shared memory environment. While locking would work too, it does not scale well and is a real pain to take care of, since it has to be re-done for every variable introduced (or locking gets painfully slow). So putting everything into atomic transactions should help the programmer to get better results quicker.

You can read about Transacional Memory here and here.

Now that Ian Murdock is working for Sun, he has the task to reform Solaris into a distribution/package based OS in contrast to the current monolithic approach.

In the interview he stressed the importance of a binary distribution channel for modern operating system environments. And the need for flexibility in choosing the appropriate parts for the job at hand.
Most of his statements are very similar to those of the Debian project. That leaves the question if the success of Debian can be repeated by a similar technological concept created by the same man.

The first release is announced for fall 2007.

Putting up the question whether to rewrite huge projects or not sounds like trying to start a flame-war, but surprisingly, the comments are quite sensible and lack absolute conviction.

If only all big questions in programming would be discussed in such humble ways!

The concept of Code Katas eluded me until now. But it is very intriguing, given one is able to practice on a regular basis.

For the record:

A Coding Dojo is a coding session centered around a programming challenge. The challenge is small in scope and often patterned after pragmatic Dave Thomas’ idea of Coding Kata. It is a regular programming class where programmers of varying skill levels meet as equals.