Ich hab meine Website gerade runderneuert. Es wird nun ikiwiki unter der Haube benutzt und alles in git verwaltet.

Dort sollen nun längere Artikel landen, aber keine unreifen Gedanken oder dummen Internetwitze. Inwieweit dieses Blog hier nun weitergeführt wird, da bin ich mir noch nicht ganz sicher.

Cloud computing is all the rage currently. Many companies try to hit that new market (Google AppEngine, Amazon AWS, Windows Azure, RightScale, Heroku, Salesforce, ...) and many people discuss about business models and platform strategies. Will there be one massive company owning the market (like Microsoft or Google) or will it be done by lots of small companies (like web hosting)?

Personally i'm more interested in the technical questions. So how will applications "in the cloud" be developed?

1. Operating system independent - since all the major OSs are not distributed, the current mechanisms probably don't work as good in the cloud. One can see how NFS breaks some assumptions about how a file system works. Just read about mbox and nfs.
   
2. Fault-tolerant - running software on an Erlang cluster is probably very "cloudy". In a distributed system it is common that nodes crash or disappear and the developer must keep this in mind.
3. Web-oriented - the biggest opportunity for cloud computing seems to be web services. The problem is mostly embarrassingly parallel, which makes it easy to distribute over a cloud. A Web interface is also the easiest entry point for end users.

As far as i know about modern software technology, there is nothing revolutionary new in the last ten to twenty years. It's all about minor adjustments and improvements of old ideas.

So what old idea is cloud computing based on? Distributed operating systems of course. They are just academic projects in most cases and didn't really make it to the market. The Amoeba OS, a project of Tanenbaum is abandoned and it's only legacy seems to a scripting language called Python.

But there is one project, which still has a somewhat active community: Plan9. There is also an commercial offspring called Inferno. I don't know why Vita Nouva doesn't market it more aggressively as a cloud computing platform. They seem to aim for the embedded market.

A problem with Inferno is probably the high entrance cost. While it has all the modern features, it is still a new operating system with a new programming language and very different from anything known to general public.

The solution: Implement a Cloud Computing platform that can be used with popular languages like Java by stealing the ideas from Plan9 and Erlang.

I'm sometimes using some diagnostic CSS, so my default CSS template includes this line

@import url(diagnosticX.css); /* Remove in production?! */

I just remove the "X" and some errors on the page get highlighted.

The comment should remind me, that this isn't meant to be in production CSS, yet i never bother to remove the line. Who cares? It points to a CSS file, which will never exist, so the browsers will ignore it anyways.

Turns out it can be harmful in some cases. One such case includes Typo3 and Safari.

If a resource doesn't exist, Typo3 returns a nicely layouted error page. Since the developers care about good CSS, they include some CSS rules in the head part of the error page. Safari trying to do its best to parse this file as CSS actually understands those CSS rules. The effect is that the non existant diagnosticX.css file does hold some layout commands and those affect the page layout.

The mistake is on Safaris side, i think, since Typo3 seems to return a 404 error page and Safari must not parse CSS files that come in with a 404 status code.

Nonetheless i change my template, so i don't forget to remove it.

@import url(diagnostic.css); /* Remove in production! */

Der LHC ist erstmal lahmgelegt. Das Half-life weniger Fiktion, als eher Vision war, sollte klar sein. Ob die Invasion bereits begonnen hat ist unklar.

Sowohl Freeman als auch der Gman sind bereits identifziert.

Hoffen wir einfach, dass die Brechstange und der Strategy Guide Freeman rechtzeitig erreichen.

Die Bahn verlangt derzeit 2,50€ Bedienzuschlag, wenn man die Fahrkarte am Schalter, statt am Automaten kauft. Aber nachdem Angela bei Hartmut angerufen hat, soll das nun wieder gekippt werden.

Neue Idee: Wir erhöhen die Preise um 2,50€ und geben am Automaten 2,50€ Ermäßigung.