Mono required by ubuntu-desktop

[Dan Kegel]

On 8/2/06, Scott James Remnant <scott at ubuntu.com> wrote:
> Now, don't get me wrong -- I'm all for eliminating wasted memory usage
> ("bloat") from anywhere it can be found.  The smaller and more pert a
> binary, the faster it's going to load and run on even the most modern
> machine.
>
> But this should not be done with unrealistic measures.

Sure, by all means, let's be realistic.

> Setting out a goal of minimum memory requirements isn't, in my view
> anyway, the right way to do this;  it's not going to be attainable, and
> even if it were on some architectures, there's no roadmap for getting
> there.

So let's pick an attainable goal: no net increase in memory requirements
from the status quo (unless the user starts some new huge app on purpose).

> The right way to do this is instead to set out to audit the memory usage
> of a certain percentage of libraries and applications, and understand
> how and why they use memory and whether that can be improved.
>...
> So my recommendation would be to ditch the "Ubuntu will boot (to what,
> btw?) in 32MB of RAM" targets and instead go with something like:
>
>  - for edgy+1 50% of libraries installed by ubuntu-desktop will be
>    audited for memory usage
>  - for edgy+2 100% of libraries will have been audited
>  - for edgy+3 50% of applications will have been audited
>  - for edgy+4 100% of applications will have been audited.

That's a lot of auditing.  How about we reduce the amount of auditing needed
by only auditing packages which are implicated in a performance / RAM regression
revealed by periodic performance tests?  e.g. we could define a small
test of tasks:
  boot to login screen
  log in
  open a document on OOo
  open a document in Abiword
  open a document in KOffice
measure them under a variety of conditions
  mem=96M, mem=192M
  ubuntu / kubuntu / xubuntu
and make sure these don't get worse.

> Also bear in mind that if an ubuntu-desktop installed application does
> require lots of memory, or is just simply too crappy to fix (yes,
> OpenOffice.org, I'm looking at you) -- we can always have a derivative
> (Xubuntu!) targeted at less capable hardware which ships something else.
>
> I don't see a problem with Ubuntu itself being targeted at computers
> purchased within the last year, provided that the usage of the resources
> is for a reason[1] and not simply wasteful.

Do you see a problem with avoiding performance / RAM requirement regressions?
- Dan