Jul 112014

As we are going to update to Centos 7 sometime in the far future (around 7.1 – 7.2, after a bit of long-term experience), we started installing some test virtual machines with the just released Centos 7.0.

We never really had trouble with Centos 6.5. Some packages are obviously very old (this can’t be avoided in a distribution with support cycles this long) and the configuration layout feels quite antiquated, but we never had any stability issues or problems with driver support in sipte of the old kernel.

I had high hopes for Centos 7, because of systemd. Due to the standardization of many config files for systemd support and due to the big investment of Red Hat into systemd, I thought RHEL and Centos 7 to become THE standard linux distribution. From what I saw however, everything is unchanged. Centos 7 very much feels like Centos 6. I guess I could have simply copied over most config files and everything would have run (systemctl has a service wrapper, the network configuration is completely unchanged, still no dhcp by default, etc.).

While this makes sense from an enterprise point of view, as you do not need to retrain your workforce, it is a bit underwhelming from an enduser point of view. This also means that there will still be a lot of unneccessary deviation in configuration management of the distributions in years to come.

 Posted by at 9:50 am
Oct 182013

Some choices Microsoft took in their way to handle EFI increase the difficulty to dual boot Windows 8 (and even more so 8.1) with Linux. Observing how these choices evolved with the release of Windows 8.1, I presume these changes were intentional to lock other operating systems out, when using Windows 8(.1). This post only concerns you, if you installed your PC in EFI mode, i.e. you use for example an EFI version of grub. It also concerns you, if you didn’t decide yet, whether to install in EFI mode or in BIOS mode. Short answer: BIOS mode.

Windows 8 introduces some new form of automatic hibernation. If you do not turn this feature off, you will not be able to dual boot with Linux. The following will happen:
After installing Windows 8 in EFI mode and then your preferred GNU/Linux in EFI mode, you will install for example grub-efi.
Check that it boots by booting into Linux (it works).
Then boot Windows 8 (also works).

On the next reboot grub will be gone forever, automatically overwritten by the Windows bootloader. I have not found official documentation as to why, but it happens every shutdown using the new hibernation feature. I guess Windows has a backup of its bootloader somewhere, which it overwrites on shutdown. An answer to the problem can for example be found on Stackoverflow.

Once you turn the hibernation feature off:

powercfg -h off
(Source: Technet)
DO NOT boot into Linux immediately, but do one more shutdown restart cycle with windows. IF you install grub immediately after and then boot windows, your BCD will become corrupted somehow (I guess Windows tries to set some bit in the bootloader, which since then got changed with grub) and you will get a completely unbootable system.
After another power cycle (trust me, you really need it!), install grub-efi as usual.

Now to the changes with Windows 8.1:
With Windows 8.1 Microsoft silently asks you to also enable UEFI Secure Boot, if you boot in EFI mode. This is done by a permanent watermark “SecureBoot isn’t configured correctly.” akin to the “Your windows license is not legitimate” in the lower right corner of the desktop. Currently the only way to remove it is similar to the way you can remove the watermark, if you have an illegal copy, by binary hacking some resource files (don’t do it). If you choose, like me, to have Secure Boot disabled out of your own free will, you will not be able to remove the watermark.

Again to summarize: Use BIOS mode or better: use Windows 7 or even better: Don’t use Windows at all.

 Posted by at 12:58 pm
Apr 172013

I didn’t now I needed it until I found it: python-sh

I allows you to do:

from sh import touch

In other words: You can just call Unix programs with a simple import. I don’t know, where to use it yet and whether it has security implications or even if error-handling is easy, but compared to the SubProcess API this seems awesome.

 Posted by at 2:56 pm
Apr 172013

This post is quite a niche post and only concerns you, if you want to do your Ph.D in the physics department of the KIT, as there are a myriad of unwritten rules. I assume the following now: You wrote your thesis (yay!) but still have to do your oral exam (defense, not yay!)

First of all, review the current Studienplan very carefully: http://www.physik.kit.edu/Studium/Studienplaene/

  1. 4 months before: If you don’t have one yet, choose a co-supervisor. Note: One of the two has to be a full professor at KIT (the University part).
  2. Also 4 months before: Plan a date for the defense with your supervisor and co-supervisor. The date has to be on a Friday within a running semester. Ask both, when you need to submit your thesis for review, in case you actually didn’t finish writing yet. And we all know you didn’t.^^
  3. Exactly 3 months before on a Friday: Be at the dean’s secretary office at about 7:45 am and wait for her to arrive. There are only three appointments per day. I was there at 8:15 am and I got the third appointment. The secretary arrived at 8:10 am.This is the reason, why you need to apply exactly 3 months before. Edit: There has been a small change for the first three and the last two appointments during the semester. For those, all requests will be collected and afterwards appointments will be “assigned”, so you might not get your favorite date. However it’s still safest to register exactly three months before. Thanks to Benny for telling me.
  4. After you got the appointment immediately request your “Privates Führungszeugnis” at the city hall (Bürgerbüro) you are registered in, also if you are a foreigner (It takes about 2 weeks to arrive). In case you live in Karlsruhe, make an appointment over here: http://www.karlsruhe.de/b4/buergerdienste/oa/bbmitte.de
  5. Immediately after, but at least 6 weeks before: Submit a written request to defend your thesis at the Studienbüro http://www.kit.edu/studieren/12703.php This should include:
    the request itself (a letter to the Dean, which says: I want to do my defense on the XX.XX.XX in the physics department),
    the Führungszeugnis,
    the written declaration that you didn’t receive any help or tools, which you didn’t disclose. I uploaded my LibreOffice template for the latter here: http://www.timo-strunk.de/files/public/fruehere_promotion.odt,
    your CV and
    a copy of your last Diploma.
    They take about two weeks to process this request and send it to the secretary in the physics department. She wants to have it at least two weeks before. Therefore 6 weeks is a good buffer.
  6. Print your thesis at least 7 times (one for you, one for each member of the commitee) and submit the one for the dean at the dean’s secretary office.
  7. Now would also be a good time to prepare your talk. You have exactly 20 minutes. Staying in time is very important here. Edit: You can nowadays choose between a PPT and a poster presentation. Do the PPT. As to the talk: No details, just concepts. A derivation of some equation is welcome on the blackboard, if it’s short.
  8. If you didn’t by now: Start studying! Consider especially topics, which are very far from your own subject. Also join the weekly colloquium as often as possible. The exam in Karlsruhe is very wide topic-wise. There are some folders describing previous exams. Study them. Some referees ask always the same, many don’t, but at least you will get an idea about how they ask, which is unlike the Diploma exams.
  9. Two weeks before you will receive a letter at home and work, if you work at the KIT, which discloses your final referees.
  10. Study their expertise and try to find overlap between your and their research. Make appointments with them to give them your thesis. They usually give you the opportunity to present your work very briefly. The first thing they look at mostly is your own publications (last page of the thesis); afterwards they ask what you did and go through your thesis like a magazine to find something they can relate to. If you present something they can relate to, they will ask questions to this if you are lucky. Some Professors also want you to prepare a specific topic or paper. This, however, didn’t happen to me.
  11. The secretary will schedule a final shake-hands with the dean before the exam. They will show you the room with the other two candidates for the day. You don’t have to prepare anything.
  12. Relax! Because if you don’t, you will inevitably fail miserably! Now sleep! Just kidding: Nobody failed the exam yet. It is grade relevant, however as I was told only for about one grade step (going down from the proposed grade, not going up). So Good Luck.

I get that these are mostly details I write about here. However they are details, which are passed from generation to generation all the time and sometimes fragments get missing. I for example didn’t know I had to send the request to do the exam through the Studienbüro. Thankfully I was still in time.


 Posted by at 1:22 pm
Jan 272012

Today there was a nice piece of news from slashdot calling for an Elsevier journal boycott: http://science.slashdot.org/story/12/01/27/1322234/scientists-organize-elsevier-boycott .

I completely agree with all the negative points about publishing in a high-cost journal; however the proposed solutions do not address the main problem. Scientists publish in these journals, not because they are good or high-price; they publish there, because of the Impact factor. If you get a good publication in a high-Impact journal, you will quite certainly get grants easily.

Publishing articles nowadays is terribly easy and does not cost a thing (arxiv); filtering and getting good referees however is not.


My solution for this would be a public network of papers, where everybody can publish, read and ‘sign’ those papers. If you agree with a paper, you put your signature under it and the worth of this paper goes up. As your ‘worth’ goes up your signature also gains in weight, when signing other papers. Every paper gets a comment section, where reviews can be written and errors pointed out.

If a well known professor therefore signs your work, others will catch up to it. A ‘good’ paper will gain in publicity quickly due to being sent around a lot. One would also need to include a system of diminishing returns, as to avoid groups signing only their own papers. Ironing out these points of abuse will be the hardest part of this system.

The specification above only consists of four to five sentences and yet I would call it much more stable and open than the currently completely anonymous reviewing system.

 Posted by at 2:08 pm
Sep 162011

Most eeePC 1215Bs have a Touchpad issue. For me the touchpad simply went into wrong directions; it could for example go up when going right or move not at all. If it’s also not working for you, you have two options. Contact ASUS, be stern, don’t let them talk you into a ‘driver issue’ and you will get it replaced.

Or you fix it yourself. Probably you will void your warranty, so do it at your own risk. The tutorial which helped me can be found here (below) LINK. It’s the picture below with the green tape. When doing it you will most probably break off one of the small hooks holding the handrest together with the rest of the laptop. This will leave a small gap.

After taping my touchpad like in the thread and breaking off one hook (also like in the thread), my touchpad is working very well now. I never thought it could work so well, seeing how it worked before.

 Posted by at 12:51 pm
Aug 052011

EFI is the ‘new’ bios proclaimed as the savior of all by Intel (among others). Currently it is not (for me). Here’s why:

The old BIOS infrastructure has a few flaws namely:

  1. It’s impossible to address (and hence boot) from devices bigger than 2TB
  2. I haven’t seen a bugfree ACPI implementation yet. (Both windows and linux usually work around it after they find it out)
  3. It’s all written in assembler, it’s not modular, so every new mainboard will have new bugs.
  4. It’s slow to boot because of one million checks.

EFI promised to solve Nr. 1,3 and 4. Seeing posts like this made me sceptical of this, but oh well; that one was Apple specific.

Lately I had to install two new laptops. First one was the Lenovo S205, second one an ASUS eeePC 1215B. Both have very nice hardware and run quite cold. Also both use EFI, however both also support emulated legacy BIOS installs. Of course with all my operating systems supporting EFI (Win7 64bit, every Linux), I went the EFI route.

The Lenovo S205 produces (on Windows) bluescreens from time to time and also always hangs on shutdown or reboot, when installed with EFI support. On Linux it installs, however the wireless card does not work correctly and also rebooting and halting does not work (freeze). EFI supports another EFI implementation for reboots and shutdowns like ACPI. Similar to ACPI the implementation seems to be very buggy on the laptop; also the wireless firmware does not seem to be 100% EFI compatible. In BIOS mode there are no problems whatsoever.

The 1215B from ASUS booted the Windows install (EFI version 310) and then produced a blue screen after copying the files. Enter EFI 401 (current one on the asus BIOS site). This one already rebooted, when the install was loading, so at least I had less work installing and reinstalling. Again: Everything works fine in BIOS emulation land.

So my opinion: EFI wants to repeat the whole mess we had and still have with ACPI. Even if the moronic and braindead implementation should work on paper, many mainboards will be plagued by crappy implementations.

Please Coreboot, save us… …and quick.

 Posted by at 5:55 pm
May 292011

Microsoft bought Skype some time ago (link). The tech sites saw mainly two things as the reason:

  • Deep Integration into Windows Phone 7 as the killer selling point for WP7
  • Converting the Skype userbase to using MS products

I think the reason MS bought Skype could be targeting the enterprise markets. One of the first decisions made after the deal was official was to shut down the Asterisk connection plugin developed by Digium. While also the linked article states that it is unknown whether Microsoft actually had something to do with this, let’s at least consider it did:

The Skype integration into Asterisk was especially important for businesses, who wanted to keep Skype out of their infrastructure. Implementing Asterisk allowed communication with the Skype userbase through a safe and trusted channel. This is now cut and within the 2 year period Asterisk will continue working, they need to switch to another architecture. Enter “Microsoft Telephony and VoIP server” (not an actual product). For most enterprises a Microsoft proprietary application tunneling Skype would be the easiest upgrade path. MS would sell a few Windows Server licenses, the Telephony Server license and maybe some support. While I also see the two points mentioned above as the most important ones, the enterprise sector shouldn’t be underestimated, especially seeing that neither Google, Microsoft and especially Facebook do not have much enterprise penetration with their current VoIP and Videochat products, while Skype has that (at least at my work).

 Posted by at 4:28 pm