Re: edubuntu

[Gavin McCullagh <gmccullagh@gmail.com>]

Hi,

sorry, I missed this until now.  As I kind of guess it's us being
referenced I'll reply.  I had intended posting something like this soon
anyway.

On Sat, 08 Jul 2006, Martin Herweg wrote:

> today i received an e-mail from a school in ireland.  they use my
> etherboot network cards for 70 ThinClients.

We're not quite that far yet.  We'll be at about 20-25 this September and
will be stepping up further (>40-50) over the year.  Getting money to
purchase things is rather awkward so it was considered easier to buy extra
cards in advance.

> For me it is no problem if someone switches from one free software to
> another free software as long as they stay with free software.

I think the same way, though I must admit I would rather not have moved
away from Debian-based stuff and this community.  I am a big fan of debian
and have been using constantly for a number of years (since potato).

> but maybe we can learn a bit from the newcomer edubuntu.

I should point out that we are not using a thin client server per
classroom setup, we're running a thin client network across the entire
school campus.  We want a PC in every classroom (with one or two additional
computer rooms, probably running Windows).  Initially thin clients will be
for the teachers use as the students mainly do ECDL and more recently the
Microsoft equivalent :-(.  We do intend to make them available to students
too but I'm not certain how when they need windows.

I guess the best thing is if I list my reasons.  Feel free to tear them
down if that is useful.  I do _not_ wish to start a flame war, but on the
other hand, I will be honest and try to explain why we moved.

1. Ubuntu has produced a clean, well-honed desktop.  Debian has not.  This
   is a combination of menu organisation, hotplug stuff, artwork
   (arguably) and other issues.  Edubuntu and Debian-edu are pretty much
   inheriting these.  This point weighs very highly.  We have teachers
   coming from the Mac and Windows platforms.  Impression really is _very_
   important.  Personally I now use Edubuntu on desktops and Debian on
   servers.

2. Edubuntu installs a good, minimal selection of software by default.  It is
   very important to us that there _not_ be two word processing packages,
   browsers, etc.  Multiple ones cause confusion and increase the support
   load.  These are non-technical people for the most part, things need to
   be kept simple.  I raised this issue on this list before.  One reply was
   that OpenOffice was not localised for language X so Kword was necessary.
   My opinion is that this does not warrant putting it in the default
   install, but I do understand the issue.  I realise I could remove stuff
   from debian-edu's selection but this causes issues with meta-packages,
   etc.  Besides, to me removing unneeded packages is rather backwards.

3. Edubuntu's regular release cycle can be seen as both a blessing and a
   curse.  It's a pain to have to do regular upgrades -- though if it's a
   clean upgrade which makes improvements I don't mind.  At the immature point
   most OSS is right now, upgrades are very important both for improved
   functionality and interoperability (eg OpenOffice 2).  The difference in
   quality between the software in Sarge and that in Dapper is substantial.
   Etch may come out by the end of the year, but if past experience is a
   guide, much of the software will already be months behind on the day of
   release and will continue to age.  You can argue that this provides
   stability and I would agree, but stabilising an immature product is not
   necessarily attractive.  As dapper is LTS we can stick with it if we
   wish, though I suspect we'll upgrade next summer.

4. I think the LTSP project is getting close to making local device (ie usb
   disk) access work on thin clients.  This is a killer feature for us.
   This really extends [3] in that we want to keep to a recent version of
   LTSP so we can have this ASAP.  I'm not clear of the details but it sort
   of appears debian-edu has forked LTSP. Edubutnu has a plan for this,
   albeit not really with a roadmap.
	https://launchpad.net/distros/ubuntu/+spec/thinclient-local-devices

5. Edubuntu's administration is _much_ simpler.  There is one network.
   There is a simple user manager under System->Administration.  The two IT
   teachers will need to learn how to deal with lts.conf but this is not a
   big deal.  WLUS is not very nice (sorry, but it isn't) and it seems
   impossible to give a teacher limited access to webmin/wlus which would
   have been very nice.  For now, we can work with one network, one set of
   IPs.  We will need a second thin client server in the coming months and at
   that point we will look again at whether we need to re-divide the network
   (I hope not with a gigabit backbone).  A third thin client server would
   likely force the issue, but that's okay.  If we get to that point, the
   project is already an overwhelming success.

The situation is not all good.  Edubuntu's disadvantages are mostly
infrastructural:

1. No centralised LDAP.  This is a very nice aspect of debian-edu.  I guess
   I can try and implement it, but I'm not as expert as I'd like to be with
   LDAP.

2. No samba setup.  I will probably have to do this myself.  The lack of
   smb-ldap will probably make this even more awkward.  Again, it's very
   nice that debian-edu provides this.

3. Little detailed thought given to scaling up to many thin client servers.
   This may become an issue in the longer term (at 3 thin clients I guess).
   I don't think this will be such a big problem as debian-edu already
   pushed us through this thought process.   

4. Edubuntu provides a lot of KDE-based edu software but is running GNOME.
   As a result, some of this software can be a bit slow to start up as I
   suspect the KDE bits all load up.

Basically the problems revolve around the desktop and are invariably
problems of debian as well as debian-edu.  Thin client desktops are the key
point of what we're trying to do.  Our teachers won't think about
scalability or infrastructure.  They'll want it to work easily and reliably
with minimal confusion and they'll be very swayed by how pretty it looks.
I have never been comfortable recommending debian as a desktop for a linux
newbie.  I've done it with both debian and ubuntu and the results have been
markedly better with the latter.

In an ideal world, I would choose edubuntu's desktop and LTSP system on
debian-edu's infrastructure.  It has been suggested to me that this can be
done, presumably using a debian-edu main server and edubuntu thin client
servers.  I may well look into this.  It seems a bit wasteful of effort
that the two projects operate so separately but that's the business of
those doing the work :-)

Anyway, I can feel blood pressure rising and eye-balls popping so I guess
I'll cower in a corner and await your responses.  Sory for the long-winded
mail,

Gavin