wesnoth 1.4

[Kevin Kofler]

Warren Togami <wtogami at ...> writes:
> I am personally disappointed that we would avoid upgrading wesnoth in
> order to maintain saved game compatibility.  I believe that maintaining
> the ability to play on the network is more important.

There are still servers for 1.2, and in fact you automatically get redirected 
to one when you try connecting to the default server with a 1.2.x version. (I 
tried it a few hours ago.) There are few people on it, that's sure. But the 
server does exist.

> 1) What about security maintenance?  A security hole could be found in
> 1.2.8 either client or server.  Will upstream continue to maintain that
> version?  If so, for how long?

That's a good question. On the other hand, security fixes can be backported, 
and there might even be other people (e.g. Debian) doing the work for us.

> 2) It was suggested in the bodhi ticket that users of older
> distributions should use a manual 3rd party repository in order to
> obtain a newer save-game incompatible version of wesnoth.  This method
> seems undesirable to me for a number of additional reasons (guaranteeing
> that users of this repo actually get updates, security considerations).

To me, it looks like the best solution. There are arguments both for upgrading 
to 1.4 and for keeping 1.2. If there's a repository on e.g. fedorapeople.org 
with 1.4, it allows users to make an informed choice.

And of course, there's also the option to upgrade to Fedora 9 which is around 
the corner, though that may be undesirable for other reasons, which is the 
point of a backport repository.

> 3) Keeping Fedora versions on older wesnoth releases might be less of a
> problem due to the only ~13 month lifecycle.  But what about wesnoth in
> EPEL?  Big can of worms.

You have to be even more careful with upgrading things in EPEL. People who use 
an enterprise distribution really don't want the software to break things under 
them.

> 4) Downloadable content (maps, campaigns, etc.) for the older version
> became abandoned and more scarce as 1.4.x supplanted 1.2.x.  New wesnoth
> users in the coming months will be increasingly frustrated that content
> they see on the websites/forums do not match what is available/usable in
> Fedora.  This increases the perception that Fedora is not properly
> maintaining wesnoth, and perhaps you want to use another distro instead.

On the other hand, there's a lot of existing content for 1.2 which users may 
have already downloaded and which will break with the upgrade. Not only the 
savegames are backwards-incompatible, but also all the downloaded content. And 
there isn't even always a 1.4 version available (and even if that was the case, 
redownloading dozens of addons is a PITA).

> There are a number of difficult drawbacks and hoops we have to jump
> through if we refuse to upgrade wesnoth to the latest stable as a matter
> of policy.  Is this refusal worth these many drawbacks?

This isn't just a matter of policy. Breaking savegames in an update to a stable 
distribution isn't something to be taken lightly. Sure, if you primarily play 
multiplayer, you'll want to always have the latest version because that's what 
most people on the multiplayer servers will be running, but if you primarily 
play campaigns, you really don't want an automated upgrade breaking all your 
savegames and all the third-party campaigns you had installed! Campaigns are 
something you can be playing for weeks. Wesnoth isn't just a multiplayer 
client!

> Perhaps we should upgrade wesnoth to the latest stable, and provide the
> current older version *somewhere else* unsupported in case people want
> to play their older save games.  The release notes of the update and
> elsewhere (wesnoth.org and fedora wiki) can mention how to downgrade and
> avoid yum upgrades.

Upgrading by default and providing the older version elsewhere is only feasible 
if they can be installed in parallel and if the new version is changed to use a 
versioned data directory so the existing savefiles will work with the compat 
package. Or at the very least the older version provided elsewhere has to use a 
higher Epoch than the official package, but IMHO that's an ugly hack.

You can't really expect end users to:
1. manually downgrade a package and
2. manually exclude the package from upgrades.

> I realize this is a balancing act, but the reasons against upgrading are
> in the minority compared to the benefits both short and long-term.

I disagree, for reasons explained above (Wesnoth isn't only networked 
multiplayer).

I'm usually in favor of upgrading applications to the latest versions (I like 
how Fedora usually does that) unless there's a good reason not to, but IMHO 
game upgrades which break savegame compatibility are such a good reason, and in 
this case, there's also the issue of downloadable content. It's not like Fedora 
9 is so far away.

        Kevin Kofler