pam_ssh and kdm

For users of ssh, we can use an ssh-agent to manage our keys and connect to other ssh servers without typing password explicitly. However, you will need to type the password twice each time you login to the system, then start an agent and use ssh-add to get the passphrase. This is inconvenient if your password for login and the key file are the same. We can change this behavior by editing pam files. Here is how I made it with a kdm login manager under FreeBSD.

Edit /etc/pam.d/kde, add lines :

auth sufficient pam_ssh.so no_warn use_first_pass
session optional pam_ssh.so
below

auth required pam_unix.so no_warn try_first_pass

Note that the order of the two auth line is important. For more details, read the manual pam_ssh.

Best download accelerator under UNIX -- prozilla

This software really works. It can automatically search mirrors for a particular software and download it from the fastest of them. It will separate the file to several parts so the download speed can get to the maximum.
I use it to download Knoppix CD which only costs me about 15 minutes. The speed is about 600Kb/s. So try it if you want to download any large open source software from the internet.

Hard disk upgraded

My hard disk has been almost full for sometime, and I have to copy data in and out from another external hard disk frequently. Yesterday, I finally decided to buy a new 80G one for my notebook. It is a Hitachi Travelstar with 5400 rpm and 8Mb cache. I spent 3 hours to setup everything, including back up my system, install the new hard disk and restore the system. The performance of the new hard disk seems quite good. I also reformatted my old hard disk and made an encrypted filesystem on it. I used freebsd 6's new method geli and wrote 2 scripts to mount and umount that filesystem. Now I'm quite happy with the result. I won't need to worry about the security of the data on my external hard disk any more.

Possible research direction

After playing with Green functions for some time, we found that the perturbation method using Feynman diagrams is quite difficult to calculate, especially for high order terms. It seems that one can use mean field theory to get some qualitative results. However, if the fluctuations locally are important, mean filed theory will fail to work. I'm not sure how it works on our problem. But my supervisor noticed that the calculation can not converge properly. Another way to solve this problem may be renormalization group method, which is claimed quite powerful. However, I have no experience on this subject, so I'm reading some papers on that which date back to 1970s.