Another idea for comments

[Emmet Hikory]

Len Ovens wrote:
> As anyone can tell UbuntuStudio is more audio oriented than any of the
> other areas we support. With the live DVD we have lost the ability for a
> user to not install workflows/metas they don't want.
<...>
> In the STRUCTURE file in our seeds there are two sets of things we could
> change. I don't know if this would have the effect I expect or not...
> maybe someone can tell me :)

    Indeed, yes, it is *very* possible to have the live session contain only
the Desktop task, and leave all the other workflows as only available to
be installed from the DVD (not requiring further download, but requiring
that the user specifically intend to install them).  As you surmised,
this is related to the ship seed: I'm likely a bit out of date, so I'm
not including specifics here, although I'm happy to help resolve the
details if there is consensus that this is the right direction (catching
up on the changes in those areas is on my TODO list anyway).

> While installing ubiquity downloads new versions of software, perhaps at
> that time the metas could be downloaded from the dvd depending on which
> is selected by the installer (the person installing not the SW :) ).

    This is less feasible: while it would be possible to copy the relevant
files to /var/cache/apt/archives, apt is going to want to verify from the
source that the local cache is correct, so it becomes very complicated
unless we're expecting the user to regularly have the DVD in the drive.
There was some related work mooted for the ARM server effort at the
Oneiric UDS, but I don't happen to know the current status of that: it
involved making a local archive available on-disk, significantly increasing
installation footprint as a trade off for reducing network dependence for
task installation.  In any case, such a thing wouldn't happen during the
package update part of the install: it would be a separate action taken
by the installer.

    On the other hand, ubiquity itself is highly modular: some other flavours
have flavour-specific frontends.  Similarly, we could modify the frontend to
*ask* the user which workflows they would like to install as one more install
screen, perhaps with a short description of each.  We could then install the
features the user desires from the DVD as part of the install phase, so that
the user need take no additional action, and has their preferred workflows
available immediately for use.

    Unfortunately, such a model has a downside: we would need to provide
a mechanism for users to install the workflows in the live session, perhaps
with additional menu items, or desktop icons, or something.  Users would
then have to unpack and install this software to their RAM-backed tmpfs,
which reduces the RAM available to run the tools.  Given that many of our
tools are highly RAM-intensive, this may not generate the best impression :)

    Another option would be to include all the workflows in the live
environment, again modify the ubiquity frontend, and uninstall all the
workflows *not* selected by the user during the software removal phase.
The main issues with this solution are that either the user might end up
missing software they expected to be there (because they failed to select
some workflow they wanted but didn't know they wanted), and that the
uninstallation phase will take a long time/waste a lot of power/${random
negative interpretation of shipping "useless" software and removing it}.

> Then also have an extra install item on the main menu with just our metas
> in it.

    While this is interesting as a means of software discovery for users
who don't have one or another workflow they might like to try already
installed, I think it is better considered entirely separately from the
question of how/what is shipped on the live DVD (although, if we can be
certain that all sensible installtion methods result in all workflows
being installed, then this becomes uninteresting, because no user would
ever use it).

    Separately from the above, as part of my catch-up reading, I thought there
were some threads about merging the live and alternate images.  While I don't
know the technical means by which this is being done, I strongly suspect that
the changes related to this effort will be a great aid in finding specific
solutions to the question of what is shipped vs. what is installed by default.
I also suspect that the mechanisms by which the various parts are identified
may be subject to change as a result of these changes: it is worth some
investigation into that effort to make sure that the latest solution is the
target for our anticipated changes (as it would be exceedingly frustrating to
generate a solution that needed significant rework prior to release).

-- 
Emmet HIKORY