Saturday, November 28, 2009

My Worst Experience with Fedora GUI

I have been upgrading my parents computer since Fedora 8 by using yum update options.

I decided that before moving to Fedora 12, I should cleanup some of the packages. I decided to use the gui - packagekit and opted for removal of development group and some language packages like arabic which my parents did not need.

Packagekit mentioned that packages depending on these will also be removed. I clicked ok. For a moment, it did seem to me that the number of packages meant for removal was larger than I had expected. Anyway, I could always fix the problem if some needed package was removed.

After 20 minutes or so, packagekit had done its job. I could find very few packages. I decided check what was installed and rpm was missing! yum was missing.

All the rpm packages were present in /var/cache but there was no way I could use them. The machine had no cd drive. Anyway, I found that I had no installation media for Fedora 11.

So, the only option which came to me - clean install Fedora 12. I downloaded the net installation cd. Removed a cd drive from another computer and installed it on my parents computer. The installation is in progress for the last 6 hours.

Dependencies - Shared libraries should be treated differently

I install a package and required dependencies are installed. The package manager may even ask me if that is OK for me. At least I agree without even looking except for the size of the download.

If I want the package, do I have a choice? How much information do I want?

I believe that shared libraries should be handled differently and should be translucent. Package management should highlight dependencies in packages other than shared libraries. We should be able to examine shared library dependencies provided we want to see details.

Treating shared libraries differently can have several advantages:
  • They could be implicitly removed if no package requires them
  • If a new package requires a different version of a shared library, it should be possible. It should not force an update of the existing library and a cascading upgrade of various packages.

Removing implicitly installed packages

It is easy to install a new package. Yum/synaptic will find the dependencies and install them.

How about removing a package?

It is hard to remove packages which were installed because of dependencies. E.g. install mplayer and dozens of packages get added. Now, remove mplayer but the packages installed implicitly remain installed.

On Fedora, package-cleanup provides some assistance but not enough.

Wednesday, November 25, 2009

Wanting to test a new version

Package management should make it easy to test some packages. E.g. I would like to try Firefox 3.6.

If I try to use the rawhide repository, it has a dependency problem. I skip packages with dependency problem. It offers to install 14 packages, upgrade 18 packages and skips installing Firefox!

I want to test Firefox, not upgrade Python from 3.6.2 to 3.6.4!

I can and do test the beta version by downloading it from the Mozilla site.

Often end-users will want to test and try packages like firefox, openoffice for small improvements which matter to them. They will not want to figure out the dependency issues of testing a Fedora 13 package on Fedora 12. Nor are they likely to search the OpenOffice or Fedora sites for the beta versions.

Since end users are likely to remain within a distribution, wider testing of common packages needs help from package managers by Fedora and Ubuntu.

Package managers should allow a test version of a common package to be installed easily without getting rid of the stable version and allow users to test and provide feedback on the features that matter to them.

Chrome OS Possibilities

Recently, I used a desktop computer on which I had no access. I could boot into the SplashTop desktop environment.

I needed to work on the net and could do my job pretty effectively. So, why should ChromeOS be only for netbooks? I can imagine it being installed on every system as a second OS by default. A pretty nice convenience.

I can imagine it being offered as a VM so that all online financial transactions may be done using a secure VM rather than windows environment. If Google does that, I am sure even Indian banks will no longer say IE only!

An Analysis of an Upgrade to Fedora 12

The upgrade to Fedora 12 required over 2gb of download. I wanted to measure some of the changes in the upgrade for my usage.

Number of packages installed increased by 131 packages. (F11:2232, F12:2363)

Some packages like the kernel and gpg-keys have multiple versions installed and some have both x64 and i386 architecture packages installed. Hence, let us consider the unique package names.

Unique package names increased by 121 packages (F11:2087, F12:2208)

26 packages present in F11 were removed, while 147 were added.

Among the common packages, 1289 had the same version in both F11 and F12. They differed in the release value.

Of these, 343 differed in the version after the 2nd decimal place and can be regarded as fairly minor version changes.

62 differed in the major version number and 380 in the first minor version, e.g. gnome desktop going for 2.26.x to 2.28.x. A number of packages, e.g. sugar packages, have a single digit version number.

How many of these 442 packages could not have been incorporated in Fedora 11?

Would delta rpm packages not be feasible across the upgrade?

Could there be better ways of managing packages which can handle additional usage patterns?

I plan to explore some simple usage patterns I have found hard to do with current schemes of distributions. (I am also planning to install Arch linux as an example of a rolling distribution to see if they meet such needs.)

Sunday, November 22, 2009

More Roads - Solving the Wrong Problem?

A few days ago, an article reported that we have attained the distinction of having the deadliest roads in the world. The only surprise has been that the statistics are not worse.

Earlier this week, I was moving around in Pune and noticed that there was traffic in all directions. Why?

Next day, I decided to visit my cousin in Delhi on my return. It took two hours in the middle of the day from the airport. On a good day, it takes 45 minutes. We were stuck at one traffic junction for 45 minutes. Why do so many people need to be on the roads?

Should we not be looking at ways to reduce the need to travel?

Children go to school quite far from their homes. A fair number of vehicles at the time of the jam were carrying school children.

It should be easy to rent a place with minimal worries about getting the place vacated when needed. Why force a tenant to vacate every 11 months. My father got his house vacated because he was a retired government servant and that also took a ridiculous number of years! My father had finally filed a court case when his (lawyer) nephew had convinced him - "Has the tenant been bitten by a mad dog that he would leave an independent house in the middle of the town to stay far and pay much more rent?"

Another Upgrade - Fedora 12

Continuing with the strategy I have used with the last few upgrades, I installed fedora-release-12-1.noarch.rpm and cleaned the /var/cache/yum directory.

Next step was
sudo yum update --skip-broken

The upgrade went through and took about 20 hours on a 512Kbps line. The broken packages seemed to have been caused by desktop-data-model, which does not appear to be present in Fedora 12.

The installation was clean after I removed desktop-data-model and updated the broken packages.

At first glance, everything seemed fine. I could start an X session from a thin computer as well.

However, Miro wouldn't work. It seemed to run but no display. I subsequently found the same issue with Nautilus. Google search found a person with a problem using Nautilus and he could overcome it by removing gstreamer-plugins-ugly and gstreamer-plugins-bad. This worked for me as well though it did not seem very logical.

I reinstalled both one at a time to find the culprit. However, there was no problem! I presume the broken package during the upgrade may have been the cause of another ir-reproducible experience.

This time I have kept a record of the rpm packages and versions before the upgrade. I hope to analyse the versions to test my belief that the upgrade effort is excessive for the benefit. We need a better way to manage packages.

If I am convinced, I will explore Arch Linux or Gentoo and switch my parents' computer as well. My parents, who are over 80, have not had any difficulty in using Fedora for their common tasks, except when I compulsively update or upgrade their system for no particular reason and break some application. I am searching for a quick undo.

Friday, November 6, 2009

Overcoming Xserver issues on Ubuntu 9.10

Ubuntu 9.04 worked fine with 2.6.31 kernel. Switching between desktop applications was quick. This gave me the confidence of trying 9.10 again.

Xserver Freezing

Although the chipset in my system was 845 and I was not facing a blank screen on booting, I decided to use the recommendation of setting i915.modeset=0 for intel chipsets in the kernel options. Now, Ubuntu 9.10 worked well. However, the X server locked up once just before the end of the day.

Monitor Resolution Issue

For getting the monitor to work at 1280x1024 resolution, I had to add a mode line in xorg.conf file.

Progress but not yet perfect.

Thursday, November 5, 2009

Getting around an Upgrade Constraint

I love to upgrade but can't compel my colleagues. So, while I use OpenOffice 3, the standard in the organisation is OO2. The documents I need to share are simple but, unfortunately, the indentation of bullets is lost when I create the document in OO3 and it is opened in OO2.

Hence, I use OO2.4 for documents I need to share with others. After installing Ubuntu 9.10, the OO2.4 installed in my personal directory would not work. Library version issue. OO2.4 needed older version of libicu libraries.

I simply copied the desired library files from the 9.04 partition and now both versions of OO work fine.

This again brings up one of the issues with current distributions. Linux allows multiple versions of libraries to exist easily. Should package managers help in keeping multiple version of some products? I believe they should. However, I also realise that it may not be easy add these capabilities into the existing distributions.

Ubuntu 9.10 - Problem with new Releases

I couldn't upgrade my Ubuntu 9.04 installation - the upgrade script wasn't on the desktop-iso I had downloaded. It wasn't critical as I had a free partition and installed Ubuntu 9.10 on that. A lucky move.

My initial reaction was very good. The 1280x1024 resolution of the monitor was not found; however, the desktop was significantly snappier on the 512 MB system. I had been expecting and looking forward to it because of the improvements in 2.6.31.

Then, I noticed that X-server would freeze occasionally. I learnt the Alt-SysRq-k combination instead of the Ctrl-Alt-Backspace. I needed to use it often. After killing the X, the screen would usually be black and I could do nothing. I had to learn the Alt-SysRq-b combination as well!

I tried to find the cause for the X locking up but could not. It may not be just the Intel driver on an old 845 graphics chipset as Google search reports problems on Nvidia and ATI chipsets as well.

By the end of the day, I had gone back to 9.04. Then I decided to copy the 2.6.31 kernel and the modules on to the 9.04 partition. I booted 9.04 with the 2.6.31 kernel from 9.10 and the system was stable for the next half hour till the end of day.

Tomorrow, I will explore that environment further and am hoping that the problem is with X. If so, I will work with the snappier 9.04 unofficial upgrade.

This confirms my view that the current mechanism for distributions is now inappropriate. It is too disruptive and, potentially, very painful.

I will find time to explore a rolling distribution Arch linux or Gentoo. My hesitation so far has been that their repositories may not be as comprehensive as Fedora or Ubuntu. Now the benefits of a rolling distribution outweigh the drawbacks.

Saturday, October 31, 2009

Side-effects - Ignorance isn't bliss

I wasn't expecting any pension and it came as a surprise that I was eligible for one. So, why am I bothered when the amount turned out to be about Rs. 1000 per month (US$22 sounds even better!).

Anyway, I decided to understand the pension scheme and its implications. Amazingly, the pension depends only on the final pensionable salary which is approximately the average salary during the last year of service and the number of years of service. The formula is

Monthly member's pension = Pensionable salary X Pensionable service / 70

A small Python program was helpful in understanding it:

start_sal = 10000
av_incr = .1 # 10%
av_return = .085 # 8.5%
pf_acct = 0
for year in range(N):

salary = start_sal*(1 + av_incr)**year
pf_acct += salary*(1 + av_return)**(N - year - 1)

pension = 12*N*salary/70. # annual
eff_ret = 100*pension/pf_acct
print N, salary/start_sal, pension, eff_ret

The legal requirement for an employer is restricted to Rs. 6500 per month. Hence, if the employer contributes 8.33% of the legal requirements, we are clearly the loser. At the end of 35 years, the pension would be equal to a little over 3% return.

However, if the average increase in pensionable salary is 10%, which is a very conservative estimate, the return is 21% and one wonders who is paying for it! This assumed the current EPF interest rate of 8.5% which may not be sustainable. The retirement salary would be 25 times the starting salary in 35 years. A very low estimate in my opinion.

LIC pays me an annuity at, I think, 6.75%. However, I have neglected the fact that the capital will disappear once recipient disappears and government contributes 1.16% to the contribution as well. This may result in a correction of about 2% in the return, hardly enough to avoid bankruptcy.

It becomes a pyramid scheme where our children or grandchildren will be stranded with a bankrupt organisation!

Still, it does not make sense that why would the government create an organisation which was destined to failure. There must have been a rationale.

Probably, the scheme was introduced when our interest rates were high. If we assume the return is 12% and the salary increase is 10%, the pension corresponds to about 12% return.

For me, the low pension was a side effect of leaving a corporate job and joining a college on a contract assignment! Since the pensionable service was the minimum 10 years, the impact is not meaningful. Still, it bothers :)

Update 25/11/2009: The pension office informed me that currently by law, the pensionable salary is limited to Rs. 6500 only. Hence, my pension is correct, but which raises the issue, why should anyone be a member of a pension scheme which will give only 3% return? Why is government not scrapping this scheme and allowing people to use the New Pension Scheme instead?

Monday, October 12, 2009

Mortgage and Reverse Mortgage

Looking at the property values, I think the best option for a youngster would be to take a mortgage as soon as he/she starts work. Keep repaying it for most of his/her working life.

Finally, take a reverse mortgage on retirement and live off the house!

Mobility and Transaction Cost

I have always preferred to stay close to where I worked although it has not always been possible. Commuting has always been a pain.

A recent news that about a third of the Indians now are migrants struck my attention. If I migrate, I should be able to sell my residence in one place and buy it in another.

Why should it be so hard?

Why can't the transaction cost be reduced? Why not charge stamp duty on the incremental value, like VAT and not on the cost of the transaction?

I would expect it to moderate prices of real estate.

May be one day our states and municipalities will change.

Saturday, October 3, 2009

Father's Autobiography Site

We added a few chapters to the site of my father's autobiography, Lahore To Delhi.

Python Programming for Friends

I was pleasantly surprised by two requests for copies of the articles I had written in Linux for You, aimed at people who may not be computer engineers.

So, finally, yesterday on Gandhi Jayanti, with a lot of help from my wife, we added the articles on

Hopefully, they will be of use to some people over the years.

Sunday, September 20, 2009

What if everyone became frugal

The recent drive of the government to encourage austerity among its leaders made me recall the project on connected mathematics and how hard it is to make sense of complex phenomena.

If there is a shortage of train seats and those who could travel by plane decide to practice austerity, what happens to the people who could have travelled by train but could not. These people will have the option of travelling very uncomfortably by bus or unreserved train. Or if the task is urgent, they can always experience the luxury of flying which they could ill afford.

It would be great to model what would happen to the society if everyone decided to adopt austerity. I still have to get around to learning and exploring netlogo.

Saturday, September 19, 2009

Risks and Moral Hazards

I was searching for some simple problems in finance to demonstrate the use of scipy for such cases. I realised that I needed to understand a lot more and decided to explore the Financial Markets Open Yale course.

The perspective of risk management was quite enlightening, e.g. a few sentences from the 3rd lecture of Prof. Shiller :

  • The problem with long-term risks, also, is that anything that we do to mitigate these risks creates moral hazard.
  • When you manage risks, you create moral hazard. That's why we need invention and theory in finance to minimize that.
  • Everything is evolving, so I'm actually presenting here our modern finance as the outgrowth of socialism, but that's not the usual way to present it.
He quotes the classic example of fire insurance - I get fire insurance on my house and so I behave badly: I deliberately burn the house down to collect on my insurance.

I have, possibly unhappily, known and accepted that people who handle money will earn a lot. Not explicitly mentioned, but I suppose it is implicit in the lecture that if a person is handling a lot of public money, better pay him well or there is the moral hazard of him cheating.

Monday, September 14, 2009

Exit Car Park, Enter Food Court

In the commercial sector 17 in Chandigarh, I often find tables and chairs in the corridors in front of inexpensive eateries. These often disappear - I suspect, when the police 'raid' the area.

It is obvious that the eateries are encroaching on public spaces but who is the real beneficiary?

I take my car and park it for a full day - for Rs. 5 (10 cents) only. Suppose a convenient part of the massive parking area is earmarked for a common food court. Each user can pay Rs. 2 for the luxury of sitting and eating for about half an hour. A person is likely to need 1 sq meter of space in comparison to at least 5 sq meters for a car. It will certainly benefit a lot more people than the car park.

LXDE on Fedora 11 Diskless Machine

I have a five year old machine without a hard disk which I boot using a liveusb.

I have been creating a custom livecd by adding a multimedia components from the rpmfusion repository. I decided to explore the option of lxde on the liveusb after reading an article on Lubuntu in Linux Magazine. I modified the fedora-livecd-multimedia.ks replacing @gnome-desktop group by @LXDE group. As before, I retained mplayer and the gstreamer bad and ugly plugins.

The creation of the livecd and the liveusb was smoothe.

Booting created one problem. Instead of gdm, slim was used as a simple login manager, which would not give me the option to switch from gnome-session to LXDE session. I just created a file /home/liveuser/.xinitrc with one line -

That was it.

Anyway, I normally use this machine in init 3 setting, and usually use it start a remote X session on my main desktop.

Icewm is restrictive. Advantages of XFCE has been diminishing as it has become more functional. So, LXDE does seem to be a very good compromise. I am now also using it as the default session for remote X sessions. The snappiness of the desktop is impressive and ideal for the work I do from the diskless machine.

Sunday, August 30, 2009

Open Source, R&D and People on Bench

R&D is a non-trivial activity. Most organisations have specialised groups working only in this area. The idea of allocating people to R&D with the proviso that the people may be pulled out any time with no notice seems absurd. That may be because of the way we are currently structured. Much like some of the banks whose branch organisation is unchanged but now give computerised passbooks. If the person responsible for printing passbooks is absent, the answer is 'sorry, can't be done, come back tomorrow'.

R&D is not merely creation of something new and different. There is also an applicative role. An organisation needs to know and understand how a new technology may be applied to a problem. Often crash programs or urgent recruitments need to be organised when a project requires the use of an unfamiliar technology or product. In this case, it is a project specific activity. On the other hand, if we recognise the possibility of some software being potentially useful and start exploring it, especially before it becomes successful, we would regard that as an R&D activity. This activity can be done in fits and starts provided there is some continuity and guidance.

An example may help. We got involved with a AJAX tool a few years ago. I had reservations about that product and would have preferred an open source alternative like Dojo. Dojo was new, with little documentation, and we knew nothing about it except my perception that it was the better option. The project time constraints did not leave any room for experimentation. We went with the required product and faced a lot of headache in trying to achieve the desired objectives, often winding up rewriting code using our own JavaScript routines. Had we written the additional code for Dojo, our name may have been in one of the source files.

An attraction of open source software is that we do not need to spend any money to buy a product. We do not have to restrict the number of products we can afford to explore. There will be no question of having to justify the cost of exploring a product. No one will say, 'By the way, what happened to the $1000 we had spent on X?”.

The missing piece in this environment is the manpower. We need to have a small core group which is passionate about open source and technically sound. The group need not have all the skills but must have the technical and people skills to guide and advise people on virtually any technology. The persons finishing a project are often bursting with ideas. There is often a feeling that we could have done better or there must be a better option. There could also be the category who have been bored stiff with the mundane work and would like a challenge. It is my belief that the people an organisation would like to retain will fall in one of these categories. All they need is for their energies to be channelled and their efforts recognised.

The next stage would be that these people could contribute to the projects. No ad or marketing rep can be as effective as contributed code to prove that an organisation can use and support a technology. But this is an incidental benefit.

Time is a consumable. Given the size of the Indian IT industry, it boggles my mind to think about the number of brain-hours that are, perhaps, not used every day. Can we afford to continue on this path?

Workforce on the Bench

I had written the following article about 2 years ago. I remembered it after watching a TED talk on motivation by Dan Pink. I have always been bothered by the motivation problem of being idle at work and the sheer waste involved. So, I decided to place it here in two parts.

A man in a village discovers that the scales in the local shop are not balanced. He computes the losses incurred by the villagers over the generations. He is so overcome by the sheer magnitude of the losses that he loses his mind. The memories of this story, I believe by Heinrich Boll, that I read over 30 years ago, came to my mind after reading about the manpower utilisation figures of some of the software companies.

Full utilisation of manpower in a service industry is not possible or even desirable. In the software industry, we know that 80% utilisation does not mean that each person is idle for about 20% of the time! If that were so, we would have implemented one of the significant attractions of working at Google. Unfortunately, I am reminded of the time when I was waiting for an assignment. It was in the stone age, without internet or even personal computers. I kept myself busy reading magazines, computer books, anything I could lay my hands on. And of course, waiting for 530 PM so that I could go home, mentally exhausted from doing nothing. Some days were particularly bad when I was asked to do some analysis urgently just when I was about to leave and had to miss my chartered bus. Had it gone on much longer, I would have lost my mind, or more likely, quit.

Another interesting example was when I was working for the Indian market and was perpetually short of resources. A colleague with whom I had worked earlier had just returned from an assignment and, well, did not like the idea of being idle. I could definitely use him even on a day by day basis but his manager turned it down. The manager's reaction, 'I would rather keep you idle for 3 months than risk delaying sending you for a project by a month.' This colleague did not wait around for much time. He switched jobs after a month.

Lesson from the Past

The siblings playing in the crematorium, the Swine flu, the Global warming remind me of the following about the Renaissance.
In the wake of the black death, reduced population left work-forces depleted: this tended, throughout Europe, to give workers more bargaining power, particularly skilled workers. This lead to a shift of power away from rulers and towards workers and merchants, particularly in smaller states (such as composed Italy at the time). Thus, regardless of its spiritual and psychic impact, the plague's economic (and consequent political) legacy may have helped set the scene for the Renaissance.

A Vision of the Future

I was at a crematorium last week. I saw 3 little sisters running around, having fun. Each time I turned around, there was one more sibling. I finally counted 7 of them from a toddler to the eldest girl who didn't seem to be more than 10.

They seemed happy, cheerful and ran around collecting stuff left behind by mourners.

Seeing so many of them depressed me. Future seemed bleak.

Monday, June 15, 2009

Converting Root Partition to ext4

Since I had updated the system to Fedora 11, the root filesystem was still ext3.
Boot directory is on a separate ext3 partition.

To get the maximum benefit of an ext4 system, I backed up the root. Reformatted it as ext4 and restored the backup.

The result - system would not boot. Booting into Ubuntu 9.04 and recreating the initrd after chroot to the Fedora 11 partition did not help.

I extracted the files in the initrd file and found that the change was simple. One line in 'init' file needed to be changed. The file system type had to be set to ext4 and the UUID changed to the current one.

mkrootdev -t ext4 -o defaults,noatime,nodiratime,ro UUID=6718a39a-2215-4e11-8716-455e7dcb11f4

Recreated the initrd file - remembering to use the cpio option -H newc

The system has been working fine.

Sunday, June 14, 2009

NetworkManager makes a difference

I finally managed to get NetworkManager to work for a fixed ip as well. It was really simple. Both system-config-network and nm-applet create the appropriate files.

Fedora includes a plugin - ifcfg-rh. This integrates the /etc/sysconfig/network-scripts with the NetworkManager settings. The line needed in ifcfg-ethX is:

Plugging/unplugging the cable enables/disables the interface.

The use of NetworkManager reduced the boot time by about 20 sec to 1 minute.

Unfortunately, too late for me to check out the same changes on Fedora 10

Saturday, June 13, 2009

Fedora 11 - the initial startup experience

The boot time is supposed to be faster. But what matters is what I would see on my system. I checked twice to confirm.

From the grub screen to the display of gdm screen with the same services:

1. Fedora 10: 1 minute 25 sec
2. Fedora 11: 1 minute 15-20 sec

Visible difference will come if the X is started asap and the services get started in the background.

Signing in to the startup sound on KDE:
1. Fedora 10: 35 or so sec
2. Fedora 11: 40 or so sec

So, the usable state of my system continues to be around 2 minutes. This may become longer whenever an fsck on a file system is needed.

For gnome, the signing in time has come down from approx 20 sec to 15 sec. Even then, the effective time for being able to use the system is over 90 sec.

Bringing up the network connections takes time. So, may be switching to network manager may shave a few more seconds.

Using icewm, the signing on time reduces to 5 sec though it takes time for the menu items to show up.

XFCE4 is usable in 10 to 15 sec.

Fedora 11 Upgrade - the unofficial way

I continue to believe in 'Fedora 9, May You Live a Hundred Years, Please!' that I wrote in Linux for You, April 2008. Any upgrade is a disruption. Still, I am addicted.

I continue my experiments with upgrading with minimal disruption - even though the elapsed time is much longer. The original source of this method:

I had used it to upgrade a system which had no CD/DVD and used the rpm's cached on a server.

The 'zeroth' step - 'yum update' to ensure that we have the latest updates.

The first step - download fedora-release from Fedora 11 repository and manually update the rpm.

This now ensures that yum will search Fedora 11 repositories. It is better that yum does a clean start. So,

$sudo yum clean all

Now, preferably in a screen session:
$sudo yum -y update
and sit back and do what you want.

Not entirely, I needed to erase some packages like 'VirtualBox' which had not been downloaded through one of the configured repositories.

I also had to erase 'Miro', 'ntp' and 'ntpdate' as the versions with the Fedora 10 update were higher than the ones in Fedora 11 at the time! 'ntp' and 'ntpdate' were installed as a part of dependency resolution. I installed Miro after the update to Fedora 11.

I restarted 'yum -y update'. Thirty hours later, 2.5 gb of data was downloade in about 2000 packages. The 'broadband' connection is 256K.

I had been hoping for a miracle. I had hoped that there would be delta-rpm packages for Fedora 10 to Fedora 11 updates! It would have reduced the time to below 10 hours of downloads. Who knows, it may happen for Fedora 12.

There was still 2 GB space available, but while running the transaction but before actually any update, yum complained that not enough space for some packages. I managed to increase the space to 3GB and the update succeeded.

The extra disk space used after the upgrade was .7gb. However, it needed just above 2 GB during the update process.

The use of screen was a life saver. There was a kernel failure and the konsole session died. However, yum update continued. The actual update took about 3 hours.

15 June: Came across this useful link

Saturday, June 6, 2009

More on Schools and Boredom

I got the motivation to post the above note after I got the following link.

So, I wasn't all that unusual in disliking school.

Schools Achieve 100% Results in Boredom

I came across the presentation by Sugata Mitra given in 2007 on the “Hole in the Wall” project (sugata_mitra_shows_how_kids_teach_themselves ). It puzzles me that we are aware of the shortage of teachers, the lack of resources and yet we do not seem to apply the lessons being learned regarding informal education, at least, not on a substantial scale.

Mark Shuttleworth made his money from technology, became the first South African in space, funded the creation of Ubuntu, a Linux distribution, and the home page of the Shuttleworth foundation quotes him, “If we are to lift Africa from her current circumstances, we will need a generation of learners that are gifted about the world in which they live, and the tools to understand and shape the world.”

Technology now offers us options which did not exist till even a decade ago. It is time that we re-think about schools, their role and the best approaches for enabling our children to learn. I regard that as important because I do not have any fond memories of schools.

As I approach the age of 60, the most painful memories are all of schools. My wife feels that cultural conflicts triggered by studying in the US in my teens is the cause. However, the bad memories go back much farther.

My first memory is that I wanted to drop out of kindergarten because of the painting activities. Children are supposed to love painting but I just couldn't paint the Indian flag. I would paint the saffron and then the white and the colours would mix and the result was terrible. Trying to draw the Ashok Charkra invariably resulted in a disaster. Why was I in a state of panic at that age that I just couldn't think?

The next memory is of trying to drop out of primary school. Having to memorise and repeat a set of seemingly random numbers - the table of 7 was horrible. It became worse when I had to memorise the table of 13. It still amazes me that I went on to study physics and mathematics. It could be that my experiences with English were even worse. I don't know if I ever succeeded in memorising 10 lines or so of "If you don't succeed at first". I wanted to drop out.

About a year after my father had been transferred to the US, we had a substitute teacher for physics one day. The topic we had been studying was conservation of momentum. The substitute teacher asked me what makes the planes fly. The answer or the absence of one still embarrasses me. How could I have been the 'best' student in physics and never even thought about its connection to the objects and events around us.

For the life of me, I cannot find a single memory of schools which I recall fondly. I cannot recall any teacher who influenced me or had a positive impact on me though there were teachers whom I liked very much. I did well in schools in spite of the mind being absent half the time. It makes me wonder if I was brighter or the other children had even more Calvin-like fantasies.

There was a sudden change when I reached college.

I wonder what was so different that I suddenly liked learning after coming to college. The main reason probably was the absence of boredom. There was no catering to the lowest common denominator. The professors were teaching what they liked. The passion showed. We were dumped with lots of information and it was our responsibility to assimilate it. Day dreaming in class was far less. The average number of hours spent on lectures was about 3 per day - 5 days a week. Scolding was subtle. "You don't seem to have liked the assignment. There were many careless spelling errors." Even today my behaviour is influenced by what I learnt in the psychology classes in college, although those subjects had nothing to do with my major - physics. Learning American history in college and coming across the critical analysis of it, especially, the impact on the local populations was really enlightening and mind blowing.

The same model may not work in schools. In fact, the above model may be inappropriate for colleges as well in today's world.

I would like to see schools offer a smorgasbord of toys which help children learn. A child may spend as much or as little time on a toy. Teachers should actively encourage children to learn from each other, rather than “do not copy” syndrome. A teacher need spend a few minutes to help each child get started – after that the learners explore independently or collaboratively. It is imperative that certification needs and learning objectives do not get entangled and confused with each other.

On second thought, I would be thrilled if schools implement any solution that eliminates boredom and tediousness from the lives of the students and teachers.

My son sent me this link and I had better get back to wasting time on Newtonian mechanics!

Fedora Live USB and Persistence

The disk of my 5 year old machine failed. I wanted to explore the possibility of using it from a 2GB flash disk. The liveusb creator of Fedora with persistence seemed very convenient. All I needed was a second machine to browse the net and use my primary desktop with ssh.

I wanted to add the usual multimedia applications and the persistence filesystem proved very flaky. Even backing up the persistence snapshot and installing rpm's - a few at a time did not work. The snapshot was repeatedly marked as bad.

Since the problem is likely to be the old hardware, a better workaround was needed. I did not wish to create a new live image from scratch.

I extracted and mounted the LiveOS image on the desktop and then add the desired packages to them. The steps were:

$ cp path-to-usb/LiveOS/squashfs.img .
$ sudo mount -oloop squashfs.img /mnt

Now extract the files from the squash file system
$ sudo cp -r /mnt live
$ sudo umount /mnt

Mount the ext3 image used by the LiveOS
$ sudo mount -o loop live/LiveOS/ext3fs.img /mnt

Add the repositories for rpmfusion as per but adding the root option.
$sudo rpm --root=/mnt -Uvh

Now, it was simple to use yum
$sudo yum -y --installroot=/mnt install gstreamer-plugins-bad gstreamer-plugins-ugly

The new squashfs can now be created

$ mksquashfs live squashfs.img

Finally, the new squashfs can be copied to the flashdisk.
As a precaution, I recreated the overlay file, e.g. a 256MB file as below

$dd if=/dev/zero of=overlay-FEDORA-6EB6-DB4C bs=1M count=256

I liked the Fedora 11 beta liveusb, but somehow, I could not get ssh -X to work with my Fedora 10 desktop. I will wait to try again but after the official release so that I can update the desktop as well.

Ubuntu uses a different mechanism for creating persistent storage on the USB disks. I plan to explore that as well and see if that is less flaky on my hardware.

Tuesday, March 24, 2009

Why Refactor Life

At the suggestion of the most significant person in my life - my wife, I created a new blog in lieu of "Growing Older, May be Wiser".

I think she does not like the idea of me growing older!

Anyway, I hope to continue to make sense of software technologies - using python whenever possible.

For the last few months, I have been planning to put my articles about using Python for different types of problems on the web. But somehow, time vanishes.

Hence, the desires:
if only I could refactor my life, improve efficiency and recover lost time

Sunday, March 22, 2009

My Father's Autobiography

Thanks to Google Apps, it took just a day to put the content in a reasonable form. It was nice to see that my father liked the site we created for his autobiography -