Jeff Spaleta wrote:
On 12/5/05, Jack Tanner <ihok@xxxxxxxxxxx> wrote:
Think "Aunt Tillie". She's got one PC, not several. She just wants it to
be safe from the baddies on the intraweb. QA of an update at a time,
with individual QA of each dependency, is not something she is going to
do. The simple desktop use case is to turn on automated nightly updates,
and to let power users and sysadmins turn those off at their own risk.
"Partial" automated updates do make sense for this scenario.
What AT wants and what is a best practise for AT's situation are very
different things.
I understand that people "desire" the ability to have reliable
automated updates that require absolutely no interaction with the
system. I don't think this project can provide that facility
"reliably" for the majority of users. And if it automated updates
can't be done reliably, then this project should instead focus on
notification of updates and user initiated update pulls.
I am not convinced that partial updates are in AT's best
interest...even though its desired.
Allowing normal people to easily ignore errors will only serve to
reduce the amount of error reporting that is occuring about package
updates. Letting users ignore errors by default even more so. If
partial updates can be accomplished with a click of a button, whatever
dialog the AT oriented tool throws up about errors becomes nothing
more than a nag dialog to click through. Turning update error dialogs
into EULA-like nag windows that AT isn't going to read is not useful
nor is it wise. If something unexpected happens AT needs to be told,
and needs to be told how to contact people who can help her
understand, report and resolve the problem. What if the update that
isn't being installed IS a security fix? Since the tools have no
ability to distinguish security updates from feature updates, how can
it be safe to encourage AT ignore the problems?
Because if given the choice to do something confusing or nothing at all
AT will do nothing at all (After all her system is working well right
now according to her right?). I think this is pretty self evident in her
lack of effort or unwillingness to enable auto updates from the
beginning. If you want to auto report such hangups as dep resolutions
then you might have a solution to your "missing data" assuming the users
agree to such a policy. We dont want to force AT to submit bugreports.
That really scares aunt tilly away from fedora.
I think we need to realize that there is a class of computer users that
just want things to work and wont take action when they dont. The best
option for these users is to provide them with at least partial updates
to do what we can for them in an automated fashion. Not everyone is a
computer enthusiast and i dont think we should ask them to become one to
run fedora.
I'm not saying that this should be silent. AT should know something is
wrong and the severity of it should increase with the amount of time
that it persists. She knows to take her car into the repair shop if the
engine light is on and i think we can manage the same effect on her
computer.
Aunt tilly is only one class of user and she doent need to be involved
in making these types of decisions. -mf
--
fedora-devel-list mailing list
fedora-devel-list@xxxxxxxxxx
https://www.redhat.com/mailman/listinfo/fedora-devel-list
