Vienna failed to migrate to GNU/Linux: why?

Vienna failed to migrate to GNU/Linux: why?


Several governments and councils reported multi-year migration plans to GNU/Linux. Free software activists praised each one of them in their blogs and commentaries. However, a few months or years on, some of those plans crumbled. Vienna is one of them. A question here begs to be answered: why did it happen? The City of Vienna made several crucial mistakes. In this article, I will list the most prominent ones.

I worked in IT for many years. I used to be the whizkid, whereas now I am just the "older guy who's seen it all" (I am getting old: I am over 30 now...). I've seen enough to make this very bold statement: "Migrating is hard. In fact, it might well be the hardest thing to do for an IT department". Starting ex-novo is easy: you can decide what technologies you will use, and work around problems and idiosyncrasies of the chosen platforms. There are very few "gotchas" when you start from scratch and have a clear list of requirements. When you have to migrate, however, you often enter a real life nightmare: some things are just not going to work.

A good migration is 1) Gradual 2) Very expensive 3) Painful. It's gradual, because it's easier to fix 10 people's problems (possibly representing a good sample of the real user base) than answering 1000 calls from users who really, really need that particular feature that takes 6 months to implement. It's expensive, because it requires both staff retraining, and structural adjustments (which are sometimes major). It's painful, because it's "change" and "change" is painful. Move house, divorce, or change job to see what I mean.

Here is a point list of what I think the mistakes were in the City of Vienna's migration. Although I am very direct, I am writing this very humbly: please keep in mind that it's always easy to point your finger at somebody after a major stuff-up, whereas a lot of us have made worse mistakes in the past.

Their own distribution? Hello?

Their first, huge mistake was their attempt to roll out their own GNU/Linux distribution called Wienux. In my opinion, this move showed a mixture of madness, boldness and sadistic intentions; or, maybe, just lack of experience. Wienux was based on Debian, which probably helped. However, anybody working on Ubuntu will tell you that even creating a "simple" fork of Debian is anything but easy; in fact, it's monumental. Issues like rolling out updates, merging back to Debian, local and upstream bug reporting, will eventually make you realise that it's a lot cheaper to pay for Vista licenses. In this case, they really had several options: they could have rolled out lots of Ubuntu desktops (if they wanted to go the easy route), or even CentOS desktops (if they wanted a stable giant), or a terminal server architecture (if they had an experienced IT department and if their infrastructure allowed it).

But developing their own GNU/Linux distribution? Please...

Their IE-dependent software

Back in 2003, Microsoft and the German Ministry for Family teamed up to create "Schlaumäuse" (which should mean "Clever Mouse", although I don't speak German), a program aimed at teaching kids how to use computers.

Keep in mind that back in 2003 Firefox didn't even exist yet. At that point, making sure that a Web site was multi-platform, or even multi-browser, was hardly a priority. Microsoft had (luckily only briefly) a stronghold on the web browser market, and they surely used it: the maker of Schlaumäuse managed to create a Frankenstein-like monster that required several IE plugins (heavens knows what for). The software maker that created Schlaumäuse hinted that while a Firefox version is in the works (planned for 2009!), the city had not offered enough incentives to speed up development. See: they didn't pay them enough to care. Also, surprise surprise, Wine would crash repeatedly while running Schlaumäuse.

We can kick and scream all we want about Schlaumäuse needing IE plugins to work. It's clearly a narrow-minded decision; however, keep in mind that the government (the German government, I assume) signed off the specs for the project, accepting its dependency to Internet Explorer. Keep in mind that the maker of Schlaumäuse very likely did things according to their internal know-how and resources, and that making something multi-anything (multi-platform, multi-browser, multi-processor) is always harder than focussing on one particular platform. Keep in mind that Microsoft backed that project, and most likely made absolute sure that IE was crucial for it to work.

In the end, it was the City ff Vienna's IT department's misjudgement that lead to the plan's failure: they didn't pay for a much needed upgrade to Firefox of Schlaumäuse (although, God knows how much it would have cost them); they trusted Wine, and obviously didn't test it enough before the migration started.

As it often happens with migrations, you never know who to blame first: the fact that Schlaumäuse is an IE kludge? Or the fact that Wine isn't advanced enough and crashes when facing IE kludges? Or that the City Of Vienna didn't test Wine enough? Or the fact that governments should never partner with convicted monopolists without specifying that the end result mustn't be bound to their platform? Or... Or... Or...

Hardware problems?

Another issue they raised was about hardware compatibility. This is at least partially a result of them creating their own GNU/Linux distribution (which, as I said earlier, was a bad move to start with). Somebody really must tell the City of Vienna's IT department that migrating to Vista is not going to make things much better. In fact, it's going to bring more hardware incompatibility problems, and more expenses. As I wrote above, migrations are often very expensive, often for unforeseeable reasons. Even if only the mice used in their existing computers are not compatible with Linux (which is unlikely, but it works as an example), they will have to spend un-budgeted money for hundreds, thousands of mice. Again, "starting small" are the keywords here.

Microsoft's pressure

There is one last point: Microsoft pressure. When Massachusetts's former CIO Peter Quinn made the bold move of starting the migration to ODF for document storage (starting a small, unstoppable revolution), Microsoft's "damage control" machine started moving. And made no enemies. They managed to get the Disability Rights groups to bash OpenOffice and praise Microsoft Office (which was absurd in itself). They put immense amounts of pressure onto Peter Quinn himself, who received unjustifiable personal attacks and who, eventually, resigned from his job.

The second the City of Vienna announced their migration to free software, Microsoft's "damage control" machine started; we will never, ever know what they did; words like "bribes", "absurd discounts", "personal pressures" come to mind, but they are just (frankly unfair) speculation.

In the end, when migrating to free software, the keyword is "low-profile", "hardly advertised", "very gradual"; this is the only way to escape Microsoft's radar. Unless the real goal of starting the migration was to get bigger discounts from Microsoft, which is always possible (although I don't think it was the case here).

Learning from your mistakes?

The City of Vienna might well be a lost cause. However, my hope is that other IT managers will eventually stumble across this article, and avoid making the same mistakes. The presence of a strong desktop distribution (Ubuntu) will hopefully simplify things. However, in the end we need to accept that most of the world will eventually start migrating to open standards and free software, and that it's going to be gradual, expensive and painful.

Welcome to hell.

Category: 

Comments

ellaguno's picture
Submitted by ellaguno on

As you said a migration is a major event, and is a complex process particularly the human change management part. We must learn on failures, when you only hear the nice part of the events it is imposible to avoid mistakes.

You mention about the German government with Vienna, wasn't it the Austria government or is this a shared project.

admin's picture
Submitted by admin on

Hi,

-------------------
You mention about the German government with Vienna, wasn't it the Austria government or is this a shared project.
--------------------

It was a German project *originally*. However, I guess it got adopted by the Gran speaking world as a whole.

A lot of information about it is in German, and I can't speak it :-D

Merc.

swiftnet's picture
Submitted by swiftnet on

I've migrated several smb office's from Windows to GNU/Linux. In every case I setup access to Windows sessions via a Terminal Service type program (RDP, NoMachine's NX, etc). A dedicated server running virtualization such as Xen, VMware, etc. would house the Windows sessions. In order to keep users from staying in the Windows sessions, Internet, email, multimedia, etc are disabled. In most instances we ran a modified RDP so only the specific applications that required Windows would open up in a Window on the Linux desktop. This method works extremely well. Yes, these shops are not '100% Microsoft Free', but they all have enjoyed cost savings and higher uptime due to the migration.

Cheers,

Alex C.

viennaguy's picture
Submitted by viennaguy on

Your article contains so many important errors, you should revise it. You should also apologize to the City of Vienna IT once you review the facts:

Vienna never created it's own distribution. They intended to do so, but in the end they rolled out basically unmodified Debian to 1000 of their 32000 workstations.

750 of those workstations are now reversed to Debian/XP (NOT Vista) Dual boot.

The "Schlaumäuse" Software was never part of the project or discussion. The real reason for the reversal can be read here: http://futurezone.orf.at/it/stories/285045/ (IT Pages of Austrian Televiosion, in German). Basically it confirms that the "Schlaumäuse" rumor is a rumor without any truth. It also confirms that the real reason is: The (non-IT) administrator for Vienna's Kindergartens ordered software ("Bi-Cube LC" from German software company iSM in Rostock), unaware there is no Linux version. iSMs CEO is quoted that iSM could have done a Linux version, given enough time and resources, as iSM does solutions on any platform.

You should be aware that a lot of the bad-mouthing of the City of Vienna's IT is coming from Green politician Marie Ringler. However, another Green politican was part of a feasibilty study 2 years ago, with the result that conversion of the City's 32000 desktops to Linux would be financially unattractive.

Please cheack your sources, there is no reason to bad-mouth the City of Vienna IT. They do an excellent job and are not platform focused. E.g., the majority of their server software is running Linux today (400 servers running Linux)

It would be fair to check your sources and post a correction and apology.
E.g., contact the City of Vienna (e.g. Mr. Peter Pflägling (pfp@adv.magwien.gv.at) who coordinates most Open Source activities for the City of Vienna and even speaks at local Linux Days)

Help spreading Linux, but help to stop FUD.
This story is FUD driven by an ambitious politician.
You can easily verify this by talking to the people involved.

admin's picture
Submitted by admin on

Hi,

1) The *intention* to create their own distribution is ludicrous enough. Reverting to Debian rather than Ubuntu for *desktop* systems is asking for trouble (which, as far as I understand, they got)

2) I talked to a couple of people in the project, and no, it wasn't just about _one_ product that was incompatible with the kindergarten's intentions, but a _series_ of pieces of software aimed at children. Wine developers are trying to help here.

3) I will leave politics out of my response

4) I will gladly run an interview with Peter Pflägling if he's available.

Bye,

Tonny Mobily

hjp's picture
Submitted by hjp on

I don't think this is ludicrous. It's what you do when you prepare to install a collection of software on more than a handful of computers.

Even if you install Windows: You select the software, prepare the desktop, create a software repository, and prepare install media. Ok, nobody in the Windows world would call this "building a windows distribution", but it's just what a linux distribution is: A collection of OS and applications ready to be installed.

If you start with a common linux distribution (which includes a lot more software than Windows) the work required may be mostly to select a reasonable default set of packages to install, and prepare a default user environment, but you may also need to include some extra software (Hey, I'm building my own Debian and Redhat packages, and I've only a few dozen servers to maintain, not 1000 desktops ...).

As for Debian vs. Ubuntu: AFAIK the Wienux project started well before Ubuntu was available. They could have switched, but migrating from Debian to Ubuntu is a migration, too - a smaller migration than from Windows to Linux, but still painful.

admin's picture
Submitted by admin on

Hi,

Sorry, but "no". Building a distribution is _not_ about picking packages. That's about 0.5% of the deal -- maybe less.

This is a NON-comprehensive list of things you need do to to *maintain* a distribution:

* Pick the packages you are going to install

* Create an installer. The installer needs to pick the right hardware, etc.

* Make security updates available to your users.

* Make feature updates available. Now... this needs to be done carefully, because theree is a rather high chance of breaking a few things. For example, newer programs might need a new library, etc. This is very time consuming, and requires you to have a bunch of people testing things beforehand.

* Have a nice system to keep up with bug reports, distro-specific questions, etc. This is _expensive_ in terms of time

I hear you say "yeah, but it's simply based on Debian". Well, Ubuntu is "simply based on Debian". And it's stull _truckloads_ of work. If you don't believe me, see what happens in the Ubuntu but report system when they merge back to Debian, about 2 months before a new release of Ubuntu is out.

There is a difference between theory and practice.The practice of building and maintaining a distribution is "a lot of work". I think the IT department in Vienna found that out... To me, what they should do is admit their mistake and switch to Ubuntu.

Merc.

craigTFD's picture
Submitted by craigTFD on

"Keep in mind that back in 2003 Firefox didn’t even exist yet."

Despite this, I was running a Red Hat distribution with the Netscape and Mozilla browsers on it back at this point in time. Since then, the code has been forked into Firefox and Seamonkey/Iceweasel/whoKnowsWhatItsNameWillBeNextWeek.

Most of the Windows world is slowly converting to OpenSource via FireFox and OpenOffice. I'm hoping now that Evolution runs well on Windows boxes, that we can get rid of Outlook (in addition to IE and Office) for them too.

It is always easy to fault the people who do something. At least they tried.

I've always rewritten the Microsoft OS before installing it. I find it gives me a better sense of what Bill and Steve had to put up with (sarcastic voice).

Without a small scale trial (i.e. lab environment) to test user requirements and operation subsystem by subsystem, any migration is going to face difficulties. They could have found out about such obstacles as "Schlaumäuse" early on, and had the option of migrating the package or migrating the environment by emulation (Terminal Server, Xen, etc.) as swiftnet mentions.

But to say that packages didn't exist, when 80% of the code is still the same...

hjp's picture
Submitted by hjp on

You don't need to create an installer if you are forking an existing distribution. The installer which already comes with it is probably good enough (Although you might want to have a look at FAI, etc.)

Security updates: If you keep resonably in sync with your parent distribution, you can mostly rely on them. You still need to keep track of any additions you made, as well as some stuff which affects you but wont be patched by the parent. But then again, not every bug will affect you.

Feature updates: Yup, happens, but again, you can mostly move together with the parent. If you need to upgrade something before the parent, it's usually only a few apps or libraries you need to upgrade - don't forget this is a corporate environment which is a lot more conservative than your average geek's home.

So, what basically remains is support: Yes, if you have 20000 Users, you have to support them. That's costly. If you have thousands of users migrating from one environment to another, you have to support them more, and that's even more expensive. But that doesn't change whether you call the environment of your users "a distribution" or not. A rose is a rose ...

This is not at all comparable to Ubuntu vs. Debian, where the gap is much wider.

Oh, and as I recently learned, they have switched to Ubuntu quite some time ago - the current distribution is based Ubuntu 7.x, an upgrade to 8.04 is planned. Unlike Munich's Limux, which is still based on Debian Sarge, with an Upgrade to Debian Etch(!) being planned.

Author information

Tony Mobily's picture

Biography

Tony is the founder and the Editor In Chief of Free Software Magazine