From: Robert Klemme <bob.news@...>
Date: 2003-11-20T23:27:21+09:00
Subject: Re: retry does not work


"T. Onoma" <transami@runbox.com> schrieb im Newsbeitrag
news:E1AMoNZ-0006k6-6g@fetch-bak.runbox.com...
> robert:
>
> thanks robert, that's basically what i decided to do. your code is nice
and clean to. i like that.

Thanks!

> but you know. it strikes me as funny that its thought a good idea to
implement nested chains of handlers for errors but not for message
passing.

The pattern usually used for distributing messages is Observer, i.e., you
have some event emitting item and others can register to receive these
events.  I'm not sure what your idea of "nested chains of handlers" is
exactly and why it is best suited for this situation.

> it seems like a natural fit to me. but when ever i bring it up i get all
sorts of nay nays. what so bad about this idea? i think it would be a very
powerful tool. i agree that resume should be automatic for this purpose,
and a seperate command should exist. i don't what to hurt the poor little
raise command.

The problem is that you wanted to (ab)use exceptions for this.  Exceptions
have a totally different purpose and don't fit well for a notification
mechanism other than exceptional notifications, i.e. situations where the
normal flow of control can't be resumed.

In fact, I don't see any reason to jump *out* of a context just to emit an
event.  Instead jumping *into* another context (by issuing a method call)
is perfectly ok.  And that's how event distribution typically works.

> i'm not that bad! ;)

Can we trust you on this...?  :-)

    robert