Hans Zobelein interviewed
A Lebanese cyber-terrorist from Munich. Hans Zobelein and his IOS – Internet Operating System Counter (newest data and analysis).
Looking at your numbers, one immediately becomes aware of the fact that the most common Top-Level-Domains – TLDs are missing completely (no com, org, net, gov, mil). Linux-fans might consider your results a kind of half-way-victory, while the accuracy of IOS still seems questionable?
Hans Zobelein: I do not want to complicate things too much, so I started IOS using the RIPE host count. In this database all European and most of the other regional domains are included, but all of the big names like .com and .us are missing. Of course I want IOS to be as complete as possible. As – until today – I did not hear of any freely available listing of the missing hosts, I decided to go out hunting for hostnames myself. I took the relatively small .edu-domain as a starter, with the rest to come when I have figured out how to do it efficiently. The IOS is a curiosity-project and not something I intend to make a living of. As IOS covers nearly all of the regional domains (except for .il), you could easily query the spread of OSes by continent. I would like to know which OS is most commonly used in the so-called ‘Third World’ e.g. in Afrika.
Hans Zobelein: This is possible of course. Following this idea, I have just put some charts on the web, telling you where certain OSes have their highest market-shares. I also consider doing some nice graphics like a colored map of the world. Relevance is certainly a matter here: if I query a domain with – say – 5 hosts and 2 of these run IRIX because the sysops are friends, it is a 40% share for this OS. You are not the only one to count computers on the Internet. How do you rate your commercial competitors? Where do you see the principal obstacles for being accurate when it comes to the total number of hosts (you need that if you want to seriously speak of market-shares)? How do you maximize the accuracy of the IOS-results?
Hans Zobelein: I can not tell you very much about the commercial counters, simply because I did not buy any of their statistics. Everyone has to use water when it comes to cooking. But I could imagine that the objectivity of a survey does not exactly suffer if it does not need to find buyers. The other big free server-statistics is done by netcraft. They are counting the Web-server-software. What we did is to compare the results of our investigations by trying to figure out which servers run on which OSes. Of course this is not easy, for a lot of Webservers are not dependent on the platform they are running on. But what you can see for sure is whether your numbers are complete nonsense or not. In my opinion the results do match sufficiently. There are other OS-sniffers like nmap. Nmap seems to be a very accurate solution. The downside is that nmap is much slower and it consumes a lot of bandwidth. Especially the last feature matters to me very much because I query 30 hosts at any given time during my IOS-runs, which would mean to spam my own Internet-connection using that program. We did a comparison between queso and nmap and found that the results were not too different. Your January-survey covers the .edu-TLD for the first time, so these results are not directly comparable to earlier IOS-runs’. Why do you still count the shift in market-shares between September and January?
Hans Zobelein: You are right in a sense, but you have to see that .edu is only 2.4% of the total number of hosts, which does not change the numbers so much. Take the Apple-count for instance: 1.672% without .edu instead of 1.776% – not a world away. Therefore the usefulness of quantisizing differences in market-share compared to three months ago outweighs the little fault in accuracy. Which are the queries to the IOS-database you want to do in the future? Who else are using your numbers and for which purposes? Do you want to commercialize IOS in the future, i.e. when the methodologic flaws are fixed?
Hans Zobelein: I want to include the .com, .us, .gov and .mil-domains (I might be a bit more careful with .mil after my experiences with .il). With my 486DX100 and some shared computing-time on a DualP75, which serves as Mail-, News- and Webserver as its main occupation, I am not exactly a superpower. The IOS-data are freely available under Stallman’s CopyLeft. But I can are you that I was glad to hear one of Silicon Valley’s big guys quote some ‘market-intelligence’-company I do not want to name publicly, that Linux is number one on Webservers. The same ‘market-intelligence’-wizards had numbers out on the Web a few weeks earlier that a completely different OS was leading with more than 50%. Then I ask myself: Where did they get their new insight? One of your worst critics on the ILUG-mailinglist (Israeli Linux User Group) accused you of causing the IOS-clash in October as a kind of marketing-strategy for your counting-effort. This seems quite improbable when you look at the surface only. In the end you were without doubt the winner of the game. The people spreading the alarm have probably lost some of their reputation. But you and the IOS are well-known now, at least when it comes to experts.
Hans Zobelein: All the information I have gathered to date clearly show that the Israeli security-consultants set off a false alarm. When the story surfaced to the Israeli press, the company seems to have tried to utilize the eyeballs for their advertizing-purposes. And it was a very nice story for the Israeli press: A server named ‘beirut.leb.net’ (.leb as in Lebanon) ‘attacks’ hundreds of Israeli hosts for three weeks. This bullet hit the shooter when after a while the people wanted to see the log-files of the ‘attack’ that simply never existed. (Frank Fremerey)