There's two cases I shift:

sub some_method {
    my $self = shift;
    my ($various, $other, $parameters) = @_;
}
[download]

and something like

sub process_list {
    my ($some, $positional, $params) = splice @_, 0, 3;

    for(@_) {
        # ...
    }
}
[download]

The first case is somewhat arbitrary and maybe a bit superstitious; I just like to consistently always shift off the $self when I'm writing OO code - something about that one parameter's significance makes it feel right to me.

The second case is plain and solid reasoning: I hate to name variables when I don't need to. Temporaries should always hold computed, not copied values, in my opinion. (You also have to work on @_ in case you wanted to modify the passed values - which of course is to be used seldomly and carefully and orthogonal to shifting.)

The second case is an exception of course though not exactly a red herring either.

So in general, I would agree; my (@variables) = @_; is preferred. There are good reasons not to, occasionally, however.

Makeshifts last the longest.