Søndag 30 juli 2006 19:46, skrev Gavin McCullagh:
> Apologies in advance for the length of this mail.
Here is my long answer. It's not always easy to present complicated
things in a short way, and it's not expected either :). When I was a
young engineer at the Norwegian telephone company a boss told me
(translated from Norwegian): You need to know a lot of words to use
less words in the right way.
> As this is probably the most important point for us, I'll just
> clarify. Although I mentioned the word hot-plug, I'm really not
> talking about USB sticks working here. I'm talking about having a
> clean, tidy desktop with simple, organised, unambiguous menus.
Ok. You want the things they do with Skolelinux in +20 secondary
schools at the region of Akershus. They have the following menu
structure (* = main menu item, - program in that menu):
- aMNS (Messenger)
- Firefox (Web browser)
- OpenOffice.org Writer
- OpenOffice.org Calc
- OpenOffice.org Draw
- OpenOffice.org Impress
- OpenOffice.org Math
- Change the screen size
- OpenOffice.org print admin
* Find files
* Personal files (Home)
The panel menu (at the bottom) has the following programs (after what
Personal files (Home), e-mail (Kmail), OpenOffice Write, desktop
chooser, task menu, sound control, screen size, clock, garbage.
With Skolelinux 1.0 we had a minimal profile for home PC-s that just
installed 600 MB with software. The amount of software was almost the
list over (with a irc-client instead of amsn). The lightweight
profile was removed in Skolelinux 2.0 because nobody asked for it. We
could probably add it as a option once more.
A couple of municipalities I know of started with 20 programs with
Skolelinux at their primary schools. After a year they increased the
number of programs to 50. That was done after requests from the
teachers. At the City of Oslo they have installed over 80 programs
with Windows at all their primary and secondary schools.
Both the region (City Council) of Akershus and the municipality of
Konsvinger was proud to use less programs in the beginning. After a
12-18 months the Kongsvinger added inn 30 more programs. Our
experience is that the schools doubles the amount of programs after a
year or so, If they got the manpower to do that.
Our experience is that teachers really lagging behind the pupils using
computers. The pupils uses chat, exchange home made drawings, pictures
etc., do simulations and uses web applications. They are producers of
information at home. The teachers stick to 2-3 programs in a Office
Suite. The teachers in general lack knowledge how to use computer
programs in teaching, and they stick to a product approach when using
A lot of teachers also uses the computers as a reward to the pupils
that have done their tasks earlier than other pupils. In Norway the
experts in using computers in education have warned the government and
the teachers that the schools don't utilises the knowledge the pupils
have at home, where the computer penetration in Norway at the pupils
home are > 92%.
So the situation we are in, is that there are different expectations
to computers to different school levels and user groups in the
schools. I agree that there is expected from the teachers that they
only need a few applications. Since nobody has really got this right,
and there is probably not an good answer, the general way to this
Reuse the "lightweight" workplace as we had from Skolelinux 1.0, and
add on to that different profiles for different levels and users in
the schools (without dependency problems). The problem we counter is
that there are a variance in what the different teachers and pupils
expect in the different school levels. Thats why the French developers
made a add on CD for Skolelinux. The Germans will probably also do
> I don't wish to harp on because I'm certainly not the first person to
> say this about debian. I'm not at all an expert in interface design
> so I would have difficulty describing in detail what needs to be
> done. However, I believe it is clear to anyone when it's done right
> and when it is not.
At the lover grades the teachers opt for KOffice, web-browser,
mediaplayer, and programs for simulations as squeakland and programs
from kdeedu, gcompris etc. The teachers will opt for a desktop
control tool for every pupil in the classroom as they've installed at
the schools in Extremadura. The 3-4 last year in the primary school
the teachers are starting introducing more complicated office
applications, even if the pupils hate it. When the pupils get to
secondary school they cut away 60% of the programs that are in use in
primary school, and focusing on web-applications, e-mail and 2-3
office applications with OpenOffice.org.
So what I'm writing is that there is not an easy formula for picking
and choosing programs. And there are even a lot of differences in the
different countries. That's why the Germans and the French developers
are making an add on CD for Skolelinux. They do that to tack into
account the different expectations from the different national
> > Thats totally true. In Norway the company InOut that sells
> > Skolelinux installations with server(s) and a lot of client's, they
> > does small adjustments to the menu, and removes and installs the
> > software the schools expect:
> Which is backwards and time-consuming.
I agree with you. It's unfortunate that it's two web browsers or two
office tools. And a lot of teachers don't know how to remove
OpenOffice.org or Koffice either. But this is also time consuming for
teachers that don't know adept, synaptic or some other packaging tool
to add programs.
If we install 20 programs, they complain that there are to few
programs. Believe me, then they opt for Microsoft Windows, and they
will not go for Skolelinux, Edubuntu or any other free software
distribution. So what we are doing is in reality competing with MS
Windows, not Edubuntu.
To win the teachers we have to demonstrate, and make it extremely easy
to access program richness, programs that do usable things in a
pedagogic context. They expect support for multimedia on the
desktop. First then the Microsoft friendly teachers and principals go
for the free software. If we start with an to simplistic desktop that
the teachers used to MS Office, they will not opt for free software.
So what we do is to show them:
- National broadcast in the web browser with sound
- USB stick just working
- Squeakland with e-toys
At a whole lot of municipalities we also set OpenOffice.org 2.0.2/3 to
save file in MS Office because there is so many pupils that uses their
parents PC at home, where the parents have a agreement with their job,
and they install MS Office. Our experience is that official
backports from Debian does the job good enough:
> I really think it's important to choose one and stick with it by
> default. Those who have different requirements should then be able
> to add extra packages. When you remove the default packages, meta
> packages go with them and upgrades become more problematic as I'm
> not really running debian-edu any more. Equally if you use
> backports and unofficial packages you also deviate from the
> supported norm.
I agree with you that this could be a feasible approach. Do you have
some suggestions how this could be administrated in install time? What
I'm saying is, how should the menu look like? In our experience the
system admin at the schools don't really know so well which choice
they should do when installing the system. How should we help them
> differ from those of other schools. We're using our thin clients
> more for teachers than students. Students are taught ms office apps
> in windows (not my decision). We need to give teachers best
> interoperability with others which to me means OpenOffice.
Our sad experience in Norway is that we also have to store file in MS
Word as a default. There are a lot of schools and some City Councils
with secondary schools that uses OpenDocument, and explain the pupils
and teachers what to expect. But a lot of schools don't have the
knowledge, the willpower or the time to explain this.
> My decision is obviously based on our needs not those of schools in
> general. This might not therefore be a good data point for a
> general decision.
Tank you for clarifying that. As you can see we probably should find
an easy way to make a skeleton profile with OpenOffice.org, e-mail
client, Firefox browser, amsn, and some additional media tools. Then
we should make it very easy under the installation to choose
additional but a full pedagogic package.
> > I've tried to promote the need for easy adjustable profiles for
> > different grade levels a couple of year now. Luckily the new user
> > admin system for Skolelinux will have a plug-in that give the
> > teacher or the schools system admin the possibility to tailor the
> > menu for pupils in different grades:
> Do you really think hiding things from the menu is the answer to
> this? Do many schools really want to use two Office suites? So much
> that you would install both by default?
My counter rhetorical question is ;)
Do you really want to exclude the needs for supporting two office
tools at schools with pupils from first to tenth grade, where the
pupils in the 1-4th grade opt for KOffice, and the pupils in the 7-9th
grade opt for OpenOffice.org?
To be more constructive:
The problem here is that one size don't fits all, and we have to work
out a solution that:
1. Makes it really easy to start with an installation that makes it
easy for decision makers in the schools to choose free software
instead of proprietary. My experience is that a bare-bone edu
install does not fix this. The teachers expect multimedia support,
USB pen, video support via the web browser, DVD support, and a lot
of fun programs to do teaching things. It's still teachers out
there that are told by the Windows Joeys that Linux is a text based
system, and don't support multimedia, games etc.
2. Make it really easy to tailor the desktop to your users without
complicated install time or post installation work. Some users want
a bare bone installation with e-mail, a browser, media-support, one
office suite. And the rest of the programs could be added
on. Others wants all the programs at once.
How should this two system install uses cases be handled?
> I don't mean I don't value CipUX. I just think it shouldn't be
> necessary to use CipUX to clean up the desktop on a default install.
I agree with you that two web browsers are unnecessary. Two office
suites are also to much for the most of the cases. It was suggested to
put in Thunderbird in Skolelinux 2.0, and remove Kmail that was the
default mail client in Skolelinux 1.0. Then I said that we could not
remove an application that were in use on half of the schools using
Skolelinux. Rest of the schools don't give any e-mail yet, or they
uses an other system e.g with their learning management system or e.g
their OpenXChange server.
> > On the technical side we probably have to clean up some
> > dependencies that is unwanted. And that will be more easy to do
> > with you helping us with a bug report Gavin.
> To be honest, I'm not sure what bug report I can write.
What about writing down that the metapackage break your installation
when you remove KOffice. And then suggest that you should remove some
doubling of programs, and make some of the pedagogical programs be
optional under install time?
> Perhaps this is correct. We shall see. In my experience, upgrades
> are fine as long as you keep the system close to defaults -- avoiding
> backports and unofficial packages. I recently had a horrible
> experience upgrading Skolelinux. I never quite figured out why but I
> suspect it had at least in part to do with a few unsupported
> backports such as OpenOffice.
As you know, Skolelinux was the project that pave the way for Custom
Debian Distributions. We had to do workarounds with the 1.0
version. We used cfengine and other mechanisms to fix default
configurations in Debian packages and the installer, just to get the
configurations right. We are talking about 17-18 services out of the
box and an architecture that meet the expectations for central
operation in municipality wide installations.
A lot of the workaround was fixed, and integrated with standard Debian
in Skolelinux 2.0. But there are still some configuration issues that
should be fixed. Thats why Petter have done a lot of work with
promoting multilevel configurations in enterprise ready software:
So to be the first with Custom Debian Distribution, our project has
paid penalty points with upgrades from a system admin point of
view. But we have also written guides how to upgrade Skolelinux
installations from version 1.0 to 2.0. There are new municipalities
that goes for Skolelinux every month and when helping each other, we
will be able to make better guides and expected improvements
concerning upgrades and to adjust the desktop.
> > Skolelinux based on Sarge is good enough on the desktop after an
> > update of OpenOffice.org to the backported 2.0.3 version, and new
> > Firefox with all the necessary media plug-in and Java installed.
> as well as the desktop clean-up done by InOut?
Well, it depends on what the schools wants. And I agree with you. The
doubling up of certain applications are not good. But on the other
hand different users have different needs. To handle the different
users need could be handled at install time or after the installation
is done. The add on strategy is tempting, but then we could lose the
sail pitch that the sale people uses at InOut. There are mainly them
who have rolled out most of the installations with Skolelinux.
The reason I needed a company as InOut to help out getting the
Skolelinux to the schools is that they can handle large requests. The
schools don't need to bother with handling the reused client hardware
where 1/2 to 2/3 is broken. With InOut they got 100% tested systems
with guaranties. As we have seen from other countries, they just dump
their useless hardware to the developing world. The machines that is
dumped don't work, and the developing world has to live with dangerous
waist they are not able to handle. In Norway reusing computers is a
industry. I've heard they will do the same in France too.
> Point taken. On the other hand, the longer a system ages, the more
> likely a sysadmin will need to resort to backports and other
> non-standard mods which are frequently the cause of the upgrade
Thats true. Thats why Skolelinux recommend the official backports. We
are also in favour the new release policy in Debian with 18 months
between every upgrade, and possibility to get a upgraded and security
maintained kernel in mid-life of a version.
That said 18-24 months between every release is not a long time in the
schools. In Norway the schools have performed national exams with
computer and Internet for three years in a row now. The Norwegian
Directorate for Education and Training is managing this for > 3000
schools. They tell that a lot of schools still uses MS Office 97. Some
have Office 6.0 and there are schools out there that still uses Excel
4.2. Fore them Skolelinux 1.0 is a really impressive upgrade. A lot of
schools had the option to stay with Windows 98 until 2008 or go for
Skolelinux. They did not have the hardware to run Windows XP.
They pupils are really used to all the instability on Windows caused
by the fact that a lot of schools don't have the manpower or knowledge
to maintain the PC's in a good way. The pupils installs seemingly cool
things that works with Internet Explorer or Windows from Internet.
And believe me, that really blows the stability. So if something is
wrongly configured with Skolelinux, they don't complain enough. We
experienced this with OpenOffice.org 1.1 and the flyspell (that makes
a read line under wrongly spelled words). It was freezing on diskless
workstations. This issue was fixed by turning swap on at the diskless
workstations. Then Linux got "more" memory, and every worked well.
Of course the pupils complains about the situation with using Windows
98, as they still do at many schools. They usually have access to
Windows XP at home at their parents computer. They expect the USB pen
to work out of the box. So the from a selling point of view, Windows
XP is the benchmark. And Skolelinux 2.0 with KDE 3.3 and the necessary
plug-ins is more than enough. The only things that has been criticised
with OpenOffice.org 1.1.3 is handling of tables inside of tables when
importing Word-formated documents. Thats fixed with OpenOffice.org
2.0, and we don't really get complaints about functionality other than
saving to OpenDocument as default. That has to be handled carefully,
because people can be very angry if thats not properly explained.
And a last remark before I sum things up a little. Much of what we does
of adjusting Skolelinux installation is exactly the post install
adjustments that schools has to do with the tool EasyUbuntu or
follow a particular recipe.
To sum this part up:
What you are not satisfied with the number of programs installed and
duplication of programs. You are suggesting a slimed down "bare bone"
installations where the programs are added in after the installation
is done. What Skolelinux and I believe Edubuntu aiming for, is to
reach all the users from the first year in primary school until the
third an last year at secondary schools.
Skolelinux also meet a lot of mistrust from teachers that are told
untrue stories about the lack of pedagogical and multimedia
applications on Linux. We have really addressed that over a period of
four year. And since 2003 we have experienced that at the marked place
when selling Skolelinux packages with server(s) and all from 30 to 200
client machines to schools. Our experience is that there has to be
full multimedia support on some of the machines with USB pen. It's
smart for a lot of schools with no awareness of OpenDocument to save as
Word. The upgrade cycles are around 2-3 years.
We have done a major study of experiences with Skolelinux and other
free software installation in 5 municipalities and city councils. It
covers planning and deployment of Skolelinux/DebianEdu that currently
includes 234 Norwegian schools, 33,000 client machines, and 101,000
pupils and teachers. It covers technical issues, economics and
organisation. There is also some feedback on how the teachers use free
software in their teaching. The Norwegian report is 64 pages. A made a
shortened version in English presented at Debconf 6:
> That's odd. Edubuntu advertise 48MB minimum.
> Minimal specs
> 233MHz with 48MB ram. 2MB video ram.
Oliver Grawert wrote (2006.06.15):
> our minimal system requs for an edubuntu server are:
> 256MB for basic server operation
> 128MB per client you attach
> 32*128+256 = 4352MB
I perceived this e-mail from Oliver wrongly, and the wrong numbers get
stuck in my head. I'm sorry for that. That was usage of memory on the
server side that was 128 MB for every client. But when it comes to
thin clients swap has to be turned on to handle less than 64 MB of RAM
with MueKow. Or else the schools could encounter instability on the
> > http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=367606
> This is good to know about, thanks. It sounds like it should affect
> both edubuntu and debian-edu?
Yes, the problems with Edubuntu and clients is the same as with
Skolelinux, and now also Fedora (K12LTSP) and the other distribution
that bases the software on MueKow. The general idea is to help each
other out with fixing this bugs, and also exchange experiences. In
that way we don't misuse our scarce resources. Then we can use more
resources to get the free software installed and used at the schools.
> At install time the desktop meta-package selection could be separated
> into "basic" (one program for each common function) and "kitchen
> sink" which would give you the current situation (KOffice &
> OpenOffice, Konqueror & Firefox, etc). This would allow one to
> maintain a standard debian-edu system without all the redundant
> packages if one wished.
Ok. I believe we should look more closely at this suggestion for
Skolelinux based on Debian Etch with getting this done at install
time. We have already done some tailoring of the packages that should
make this more easy.
> Properly cleaned up menus. The _standard_ program menu should have
> two levels only: a set of categories and the programs within that
> category directly underneath. There should be no "debian" menus. I
> imagine this is not easily done though as it's really a debian issue,
> not debian-edu as such.
Thats right, and a bit irritating if you ask me. But I've had a look
at Debian Etch and things have improved both with GNOME and KDE. I
believe some of the Debian-ers have been inspired by the (K)ubuntu
clean desktop effort.
> Diskless workstations (half-thick) out of the box would be nice.
Thats the goal, with sound support, USB and other nice things.
> I've worked with Moodle a fair bit in the past. It's a nice system.
> I presume authentication would be done against debian-edu's LDAP
Yepp. Thats already in beta. The developers has worked their but of to
do the LDAP integration out of the box.
> One of the issues I would expect would be how to get all of
> the students onto the correct course pages (a problem I went over in
> a 5000 student college). It's likely that every school's way of doing
> this would be different, but it probably comes down to something like
> 1. Use school management system to export studentid:classid links to
> CSV files.
> 2. Import CSV files into Moodle (and possibly SchoolTool, etc).
The are different LDAP schemas in different places in Europe. In
Scandinavia they have agreed on a standardise format for this. In
Germany they have some other mandatory fields. In France they have
other schemas. So we have worked on making this as fluid as possible
so that the Skolelinux LDAP schema don't restrict anyone, and are
> I did this work in Griffith College Dublin and we made it available.
> It's not terribly pretty, partially because it needed to solve a lot
> of awkward problems specific to that college.
> I recall thinking at the time that it would be really nice if the
> school had an LDAP or SQL database which contained student class
> enrolments as well as account details, Moodle, SchoolTool and others
> could perhaps read these directly. This is probably quite a complex
> job though.
Yes. This is the general concept. The German and French developers
have already adding this things into the Debian repository. We will
most likely put this things into the first Skolelinux CD because the
Moodle LDAP integration needs a little bit of tinkering after the
deb-packages are installed.