New project idea - kde package manager for Mobile

[Laszlo Papp]

Hello,

Let me answer the remarks sequantially:

@Daniel
> I think that is not a good idea not to use PackageKit since it is part
> of Meego Architeture [1]. So if you don't use it you won't have any
> guarantee that your software will work on any Meego compliant OS.

It is not true since the rpmlib would be the backend and that should
work in any case.
That is also the part of the Meego architecture and on the other hand
I do not know how long Nokia will keep this not Qt frontend.
There have been a Qt frontend development for Maemo, but not that I
know about Meego. Essentially I would like to start very similar that
Nokia started for Maemo.

> KDE already has a PackageKit and its actively developed [2]. I think
> you can try to contact the main KPackageKit developer and propose him
> to join the development and start to create a better interface for
> mobile devices.
> [1] http://meego.com/developers/meego-architecture/data-management#packagekit
> [2] http://kde-apps.org/content/show.php/KPackageKit?content=84745

PackageKit is not that was I think it is good. It depends on too much
libraries unneccesarily and if you liked to have a pure Qt (or KDE)
frontend I do not think it needs to
depend on those libraries. The main point is that I would not like to
have a pluggable package manager, just really a very easy one for fun
:=)

@Chani
> why should kde have its own meego package manager, exactly?
Because there are KDE users liking KDE.

> what benefit would it bring over a qt one?
There is no Qt one. There is some closed source for Maemo currently,
but I cannot help there accordingly.

@Sebastian
> Moreover, the arguments against current solutions sound pretty
> artificial ("too bloated", "plugins for backends, therefore not light-
> weight"). A lot of thought and work has gone into these solutions, and
> they were written by people who know what their doing. From my
> experience, having plugins for different backends eases good design, and
> introduces only very little run-time overhead (while making the whole
> thing way more flexible and often future-proof).

I mentioned above I would like to create a really easy, just for Meego
and rpm, nothing more and just for fun.
I have never said they do not know what they are doing, I have just
said the purpose was not that to have it on mobile when they started
it.

> I know it's always tempting to start a new project instead of re-using
> already written code, but in many cases, that just leads to another dead
> body in the cupboard on the development side, and one more half-assed
> solution that doesn't cut it and won't for a while. We've had that more
> than once already.

I agree with that at some point. But you cannot say it in general, it
is not a bad thing to start a project universally.
Very basic example is the distribution market. There are a lot of
separate distributions working fine separately.

> I'd recommend against re-writing a package manager. You'd not be the
> first one to do it, and it's quite a lot of work to get right. It would
> be a lot more benefitial to the whole of KDE to improve existing
> solutions, possibly by benchmarking them and looking where exactly there
> are possible performance bottlenecks (you'll likely run into the same
> class of problems while writing your own package manager, anyway), and
> spending time on UI-love for the existing solutions -- possibly by
> introducing a mobile UI.
>
> The major problem of today's package managers is not that they're
> bloated, or not light-weight enough (whatever that might mean), but that
> their UIs are too complicated. Ideally, we'd have something with the
> ease of use of Android's Market (just as an example), but with the power
> of Linux packaging systems.

I agree with that, but I still think that existing alternatives are
not scaled for mobile usage.
On the other hand the existing ones are larger projects than just
doing benchmarks and maybe harder extend them with a module than
writing a really very simple backend dropping the pluggable feature
and other fancy stuff that does need to be there, at least for me.

> BTW, seeing you're from the archlinux camp, did you talk to Dario
> Freddi? AFAIK he's been working on package management UIs a while ago,
> he'll very likely have useful input for you.

Yeah, I know him, he did a great job in the shaman project, but the
basic issue is that I do not know whether I need that complexity. If
not, I am not sure the existing project truncation would not finish
there what the simple and easy front is about. In that case it is
easier to write it from scratch than understand a whole project and
truncate the unneccesary features on mobile.

Best Regards,
Laszlo Papp