I think there is a point in ditto's words, but this comment should be taken into account: http://bbs.archlinux.org/viewtopic.php?p=200629#200629 too.

Here's my opinion:
-A and -U are getting to the point where the distinction is a bit silly. I'm all for the -I/--install change, as it makes more sense to me.

That said, removing -A and -U at this point in time, so abruptly, is probably a bad move. So I suggest the following:

Add -I/--install. The -A and -U options will simply call that and output a little warning that the option is old / deprecated / being removed / whatever, and then pass all params to -I (exactly the way -A does not, except with a little message).

I'm with phrakture on this one; I think -I makes a lot more sense for the operation it performs (for downgrading a package, does -U/--upgrade really make sense?).

IMO it is nonsense to introduce a new option, keep the old ones for backwards compatibility and then have THREE options doing exactly the same. Why can't we just keep -U: everyone uses it anyway, everyone is used to it, so why change it? Just keep -U and let -A stay for backwards compatibility (doing the same as -U) and then only have TWO options doing the same.

About making sense: It doesn't really matter what it means, how much sense does -S make? It only matters what people are used to, nobody will think about what -U once meant.

Downgrading a package is an uncommon task, I'm not convinced that the name needs to cover this action. Who wouldn't be able to figure out that -U allows downgrading?

-I has problems of its own. For example, if I have foobar-2.0 installed and then I pacman -I foobar-2.1, I'd expect to have both versions installed. In fact, what I did here was upgrade, or installed AND removed. And, as a new user, I would try to use -I instead of -S to install all of my packages. I mean, I'm supposed to type pacman --sync to install a package from the repo and pacman --install to install locally? Where is the logic in that? (And finally, "I" in general is a bad argument because it looks like "l" :p).

There is no correct term that will cover every use case here. I personally think that -U covers the most common action action (upgrading), doesn't confuse the user with when they should use it versus -S, and is already something that 99% of users are familiar with. It's not broken so don't fix it, especially when the fix is no less confusing than current behavior.

re-opened and removed from 3.0 bug catcher

my vote: -A is needless (some hidden things: its checkdeps is a bit faster, since there is no need to check requiredbys)

Dan did some work on removing the -A option :
http://code.toofishes.net/gitweb.cgi?p=pacman.git;a=shortlog;h=refs/heads/oven
Afaik, the remaining problems are :
1) clean libalpm too
2) check the add pactest to see if they have an upgrade equivalent. If not, convert them to upgrade (instead of removing them all)

I believe that help is highly welcome on any of these two points, since there was no progress for quite some times now.

1.) I think, that (after new parameters of checkdeps) TRANS_TYPE_ADD is formal in alpm, interestingly its only usage is in sortbydeps (as "parameter", which can be replaced by something like '@param reversed').
2.) OFF: add013.py seems to be a bit strange to me.