Re: [boost] Boost.Threads Locking Delima

From: Pete Becker (petebecker_at_[hidden])
Date: 2002-05-17 12:46:27

At 12:22 PM 5/17/2002 -0500, William E. Kempf wrote:
>----- Original Message -----
>From: "Pete Becker" <petebecker_at_[hidden]>
>
>
> > At 11:25 AM 5/17/2002 -0500, William E. Kempf wrote:
> >
> > >The problem is, as with auto_ptr, newbies
> > >don't understand the rules of move semantics so *they* are surprised.
>So,
> > >yes, I agree this is a major drawback to this solution
> >
> > It's only a major drawback if it's important to make mult-threaded
> > programming easy for newbies. Given the current state of technology that's
> > an impossible task, and any threading package that suggests that it does
> > this (in particular, java.lang.Thread) does them a disservice.
> >
> > Writing robust, fast multi-threaded applications requires expert design
>and
> > expert implementation oversight. The target audience for a threading
> > library should be experts, not newbies.
>
>A few thoughts on this:
>
>1)  In order to become an expert you have to start as a newbie.
>2)  In the real world applications are written daily by non-experts that run
>in critical systems and use MT techniques.
>3)  The goal isn't to make a library that totally insulates newbies from
>learning to be experts, but to illustrate (for learning purposes) and
>prevent the most common mistakes made (which aren't always made by newbies
>alone when it comes to MT programming).

I don't think it's possible to design a threading library today that's 
reasonably safe for use by beginners but powerful enough to be truly 
useful, despite twenty years of attempts to find multi-threaded programming 
models that work. Time spent trying to do that could be better spent elsewhere.

>In any event, I can't tell from this post whether or not you favor move
>semantics here.  That's probably not the point of your post, but I'd still
>like to hear it.

I don't know yet.

        -- Pete

Dinkumware, Ltd.
"Genuine Software"