I feel that scoping is orthogonal to comprehensions. Mixing them together results in special cases. We see now in for-comprehensions that certain syntactic constructs are illegal, e.g. you can’t do:
for {
import Foo._
a <- b
c <- d
} yield f(a, c)
In most cases, yes. But I still occasionally run into the case of wanting to stick a (collections.breakout) at the end of a comprehension so that the types get massaged into the correct shape.
Some combination of SIP-12 (as in dotty), changing the parse rules to use those of better-monadic-for (is this in dotty yet?), and the 2.13 collections rewrite should make this problem go away
“It’s just syntax” cuts both ways.
scala 2.13.0-M5> for (i <- Option(42) ; i <- Option(3)) yield i
res0: Option[Int] = Some(3)
Probably we lend too much weight to that slogan and also the word “orthogonal”.
No, and this won’t happen until someone writes a SIP for it (and other changes to for-comprehension that can achieve consensus), we’re waiting for the community to do that.
Sorry, I don’t understand your point 
I meant that for translation obscures scoping and is not an orthogonal concern.
Rereading, I see your point was that it should be orthogonal, and that a brace should just introduce a block like anywhere else.
Probably my observation makes more sense as a response to the previous implication that existing syntax expresses scopes clearly.
The topic summary says it has “an estimated read time of 16 minutes .” But that doesn’t include re-reading for context.
Well, say goodbye to that with Scala 2.13 
I already went through my code base and removed in several dozen places … 