Re: [Gnumed-devel] automatic bug report redirection (was: GNUmed & blueprints)

[Karsten Hilbert]

> > The point is that you are against the need to require from end-user to
> > use gpg-signed email in order to send automatic report.
That much is true because it simply won't happen. Believing otherwise is 
dreaming.

> 1. Most users are accustomed to create account and login to some website.
> Only few know how to get GPG key and know how to configure their mail
> client.
I agree with this analysis.

> 2. I wonder how to do it in most popular Windows mail clients - Outlook
> and Outlook Express, it should be possible, but I'm not sure.
It isn't really that difficult but still people won't do it just because then 
they
*could* report a bug with GNUmed.

> > Karsten> We didn't get any reports when we DIDN'T require people to
> > Karsten> register so why would we get reports when we do ?
> (...)
> > Karsten> It is reasonable but most people simply won't do it. The world
> > Karsten> isn't exactly filled with "reasonable people".
> > 
> > Well, we can choose to give support to "reasonable people" who care a
> > bit abut the project at least to the extent to provide nice bug report
> > to devs.
> My thinking goes this way: if someone doesn't even care to register to
> submit a bug, he is not seriously interested in using the software anyway, so
> we needn't care.

While I agree with the analysis -- to tell you the truth: I don't care one whit
whether *people* are reasonable or not. All I care about is extracting that
bug report from them.

If that involves looking at a "trivial" report more than once I don't mind
much at the current rate. It tends to be caused by solvable problems.

> > Karsten> I am not at all opposed to dependancies.
> > 
> > Nice to hear.
> I woudn't be so happy.  You *must* have time to maintain dependencies, and
> packages live their own lifes. Who monitors the dependencies now?
I do. The solution is easy:

- everyone and their aunt can have a Debian/Testing system for free
- we declare Debian/Testing to be our reference platform
- which means we claim that everything works fine on that system
- if that claim doesn't hold we fix it
- Debian makes dependancies easy
- dependancies are carefully chosen to either exist on other platforms
  or to have equivalents which are then detected in the code
- taking care of tracking dependancies for platforms other than
  Debian/Testing lies in the responsibility of the respective package
  maintainers
- so far this has never been a problem

> Recently
> I mentioned dependency of older version of wxPython, not overriden by
> newer version installed. (no offense, just an example).
No offense taken but that wasn't even a bug. It was just as it was.
GNUmed depended on wxPython 2.6. It doesn't matter whether 2.8
is available. Both are intended to be installed side by side if necessary and
Debian supports that just fine.

> The more dependencies, the more difficult it is to install software in
> environments different from our reference OS (debian stable). Think about
> Windows users.
While true I don't believe in this as a valid argument against making
software modular. We can surely re-implement Sane or Gnuplot but it
will be of exceptionally low quality.

If a Windows user has trouble fulfilling dependancies they have
no business setting up a practice management system. They should pay
you, Jerzy, or someone with similar expertise to do it for them.

Karsten
-- 
Ist Ihr Browser Vista-kompatibel? Jetzt die neuesten 
Browser-Versionen downloaden: http://www.gmx.net/de/go/browser