Re: dvd space issues
[Holger Levsen]
> Right. I thought the benefits were obvious, just like you probably
> did, when you changed the profile question yesterday, to display the
> expert options by default. _That_ is a change everybody will notice
> and will put cognitive strain on everybody, as there are now more
> options...
Nope, the benefits of adding an extra option to the profile question
are not obvious to me, but the effort involved in undoing the change
is negligible, and no other parts in Debian Edu is affected. Because
of this, I decided to give it a go, to get some feedback. I suspect
it will be undone before Squeeze, unless the benefits are obvious to
more than just me. It is a three line change in the
debian-edu-profile shell script.
> I havent seen the evalutation for that.
I wrote my evaluation to you on IRC when we discussed such change
earlier. You even replied to it, so I thought you had seen it.
For the rest of you, here are my pros and cons:
- Adding more options to the profile question increases the
congnitive strain of people installing Debian Edu, and increases
the chance of selecting the wrong option.
- Adding the minimal profile option to the default profile question
make it more visible, and allow more people to be aware of the
alternative and probably use it where it make sense instead of
installing Debian because they are not aware of the Minimal option.
- Changing profile question affect documentation writing, which have
not started yet.
I am not sure if the increased visibility outweights the increased
cognitive strain.
> And if the evaluation of this is happening right now, I wonder why
> we dont do the same for the DVD split: just do it and then see how
> it works out.
The DVD split is sligtly more work to undo (more editing of the build
system - probably have to edit 5-10 files), and affect a lot of
different parts of Debian Edu. The package profiles and the
installation set are both affected if we split the DVD and start using
the extra space, and undoing such change will be very hard. Providing
more images affect the DVD distributors (Jim and others) and the
sysadmins in schools that need to figure out which DVD they should use
when they want to install Debian Edu. Because the impact is high and
reversing the change is hard, I believe we should be more careful with
a DVD split than a change in the default template to show during
installation.
> combined approach:
> - useful for people installing both amd64+i386 edu machines
> - less packages on the DVD, thus different installs for DVD+CD for everyone
> - CD space issues, thus we install a kernel by default, which can only deal
> with a single core and 1GiB RAM maximum.
Also:
- Easy for distributors, only one DVD image to ship. No need to
ask people what architecture they need, nor any change to send the
wrong image. Half the production cost for a complete set of DVDs
(as there is only one).
- Easy for sysadmins, who only need one DVD, and do not need to keep
track of several DVDs, one for each arch.
- Less to archive in the project than with two DVDs per release, so
we save some disk space on the archive disk.
- Allow us to install i386 LTSP environment on amd64 servers.
Changing this require code changes to debian-edu-config, or keeping
i386 on the amd64 DVD.
> splitted DVDs
> - less useful for people installing amd64, which according to
> popcon.skolelinux.org are 5% of our users
Is this because the amd64 servers installed from dVD would have to get
amd64 LTSP chroots, or something else? Ie why is it "less useful"?
> - 686 kernel on the DVD, supporting multiple CPU cores and more then
> 1 GiB RAM
This is only a benefit if we fail to drop packages we do not really
want to install by default. If for example blender is removed, (or
LTSP is changed to install i686 kernel), we can get the same advantage with the combined approach.
> - amd64 DVD still exists, it's just an extra download for those few
> who need it. 5% of our users suffer a bit, for 95% having real
> benefits.
What do you mean? Why would amd64 users need an extra download? I
would expect amd64 users download one DVD and i386 users download
another DVD, and thus no "extra download" for any of them.
> To me it is+was obvious that the benefits outweight the costs.
Did you take into account the logistical advantages of a combined DVD
that were missing in your list?
Happy hacking,
--
Petter Reinholdtsen
Reply to: