Comment 245 for bug 173890

> And better than both would be a warning of some kind that informs of the
> situation, but still allows the user to override the developer's
> decision to prevent the installation of software he wants.

Unstable releases are a feature of every distribution I've heard of,
including Ubuntu.

> Implausible, since macromedia's main testing environment would obviously
> be firefox.
>
> Again, put the user in control of the situation, not the packager.

If you put the user in control of the situation, you get source code.
The very reason we don't have this is because the user doesn't want to
be in (full) control, the user wants someone else to fully test the
latest combinations and make sure it all works great. Then get the
canned packages. Do you want to patch your own kernel for exploits, or
let someone else patch them for you and ensure that they work first? Do
you want to install all your own software, or let other people see if
it's stable first?

I fully believe that this update should have been in unstable
immediately, for the reason you mentioned. But in general, allowing
opportunities for breakage just isn't an option. This fix would have
taken hours if it wasn't found to destroy Konqueror. However, the MD5
problem gave the devs time to figure out how to proceed. It also only
affected new installations, not existing ones. Stopping to evaluate
what to do about an upgrade that left some people high and dry was the
right answer. The right answer, furthermore, was to give a workaround
to the small percentage of users and let the vast majority have
workability again.

This bug was handled correctly (though I would have liked to see a
decision faster). Keeping the same mechanisms will yield correct
operation in the future.