Goals for Fedora Workstation upgrades

Josh Boyer jwboyer at fedoraproject.org
Sat Oct 4 00:58:30 UTC 2014


On Fri, Oct 3, 2014 at 8:24 PM, Florian Müllner <fmuellner at gnome.org> wrote:
> On Fri, Oct 3, 2014 at 11:10 PM, Josh Boyer <jwboyer at fedoraproject.org> wrote:
>> Right.  It's the "we don't know" part that makes it unacceptable.  If
>> we've done a good job at picking defaults, then we're going to assume
>> the users are actually using them.  If they aren't we have no way of
>> telling so it's not safe to just remove the application.
>
> Considering that removing an application does not affect any user data
> or configuration associated with said application, I don't agree that
> this is unacceptable or unsafe - users that don't like the new default
> can easily revert to the old one. When they use search to launch the

Um... except the cases where the new application keeps it's own
directory of caches and such.  Think of evolution vs. thunderbird.
I'd be really upset if an upgrade resulted in me going to start my
email and having to either a) redownload all the headers and messages
into a new cache or b) dig around for an app in the software center
that I was perfectly happy using.

> application, it's actually little more than an extra click - instead
> of the application, they get its details page in Software which allows
> them to install it. There's obviously some disruption and

Not everyone user search to start apps.  Many people use the favorites
list and they'll be irritated it wasn't there.

> inconvenience in the (one time) extra step, but it hardly classifies
> as breakage in my opinion - in particular as I would expect many

You've used Fedora before, right?  It's not just "one click".  It's more like:

1) look in favorites and notice it's gone
2) search for it while pissed off it's not in favorites
3) get even more irritated it was removed and you have to install it again
4) wait for all the metadata to download
5) have PK tell you there are 300M worth of other updates to apply
6) decided to ignore that and just install your app (and the update
dependencies it has now)
7) wait for it to install
8) manually add it back to favorites
9) finally start your task

and then possibly

10) remove the new default app you didn't want in the first place

> (most?) users to follow the new default (after all there's a reason
> for changing the default - it is supposed to better than the old one,
> at least for most users).
> I also think we should not ignore the impact that piling up stuff the
> user never asked for over time has. Technically removing an unwanted

Ok, but now you're back to arguing for swapping things out that were
never used.  Which is fine.  That isn't the concern.  If it isnt'
used, and we can tell that, then there isn't really an issue with
replacing it.

For the cases where we switch a default app, I think it might be
possible to notify the user _before_ the upgrade via the upgrade tool.
E.g. "Fedora 22 uses an amazing new IRC client that has built in emoji
support as the default for IRC. Would you like to use this instead of
the old default that has no cute picture of a steaming pile of poo?"
We can migrate data, etc before hand as well.  Then it's no longer a
surprise, and people are less pissed off.

(Clearly my example is poor and sarcastic, but you get my point.  It's
Friday evening, allow me some fun.)

> applications is not any more effort than installing a wanted one, but
> there's a huge difference between cleaning up other people's mess and
> adding something you are looking forward to using.
>
> Picking the default email client as example, assume we change it from
> Evolution to Thunderbird at some point, and to Geary some releases
> later. Following Owen's proposal,
>   (1) users who are happy with the default (or don't use any email
> client) end up with one client
>   (2) users who prefer one of the old defaults (either Evo or
> Thunderbird) end up with two clients
>   (3) users who prefer some other email client (say KMail) end up with
> two clients as well
>   (4) users who prefer to stick with Evo when we change the default to
> Thunderbird, but then
>        adopt Thunderbird by the time we switch to Geary end up with
> three clients

They've made choices to use those email clients.  We don't get to
unilaterally undo their choices just because we think we know they'll
like the new thing better.  (and in your cases 3/4 they could have
easily removed Evo when they switched to Thunderbird or KMail).

> I only see a somewhat reasonable justification for installing three
> email clients in case (4), I'd put all the other cases somewhere
> between "a bit dodgy" and "extremely messy" - in particular installing
> four clients for users that only ever used a single one (case (3)) is
> pretty crazy. If this was done to prevent irreparable damage to case
> (2) users, that'd be an unfortunate but necessary drawback. But to
> save some users a couple of minutes on updates?

No, see below.  It's not just about the updates.  It's about the fact
that, however well intentioned, the distro made a choice that the user
wasn't expecting it to make and removed something that was in use
without telling them about it.

>> And if we switch applications for a particular task, it
>> should be done with great care and planning to minimize any impact on
>> a user's workflow.
>
> Very much agreed. Ideally the new default would pick up the most
> important configuration from the old one to make the transition as
> painless as possible (to stick with the email example: import existing
> accounts so users don't have to spend time repeating a boring setup
> they already did in the past). At the very least the change should be
> well documented, including how to get back to the old default if we
> end up removing it.

It's not just accounts.  It's metadata, it's keyboard shortcuts, it's
menu layouts.  It is the entire UX.  Every time you change how someone
does a task they were already doing, that person will either grumble
about it and learn the new thing or they'll get irritated and use the
old thing.  Do that enough times though and those grumbles become more
than just grumbles and they start using something else entirely.

Now, I think we probably mostly agree that if we switch a default app
for another, those apps are (hopefully) likely to be very similar.
The transition curve should be very small, and that's something we
should be considering before we make the switch to begin with.  These
cases will probably be fairly limited as well.  However, I really do
not want to operate under the assumption that we can blindly switch
applications out from under the users.

josh


More information about the desktop mailing list