From: Charles Oliver Nutter <charles.nutter@...>
Date: 2007-12-08T12:52:02+09:00
Subject: Re: Clarification of retry change

Gary Wright wrote:
> If you get rid of the iterator-retry semantics then it seems like the 
> rescue-retry
> semantics are quite understandable within a block within a rescue 
> clause.  Forbidding
> retry in a block in this case seems too extreme.  Perhaps there is still
> an implementation concern?

I won't try to make an implementation case, though there are complications.

More important perhaps is that retry would change from one behavior in 
1.8 (re-executing the call that takes the block that has the retry) to 
another (retry acting as though it's a "rescue retry" when produced 
within a block within a rescue). So it actually both removes the 
previous behavior and adds new behavior that so far is a totally 
hypothetical "what if" case. So it seems like just disabling it in this 
context would be the better option.

What do you think Matz?

- Charlie