Re: [Boost-users] [Join] Simplified concept introduction?

Subject: Re: [Boost-users] [Join] Simplified concept introduction?
From: Bill Buklis (boostusr_at_[hidden])
Date: 2011-03-09 15:29:47

On 3/9/2011 2:15 PM, Nat Linden wrote:
> I saw the Join library on the review queue and am somewhat curious.
> But I find the terminology a bit off-putting: ports, chords, joints.
> I've read the Introduction, the discussion on Join concurrency and the
> comparison with Futures/Promises. I know this sounds awfully dumb, but
> I still don't understand the moving parts or how they relate to each
> other. The Introduction leaves the impression that I could only
> understand (would only be interested) if I were already a fan of
> JoCaml or C(omega), neither of which I'd ever heard of before.
>
> The library author might reasonably assert that I should read all the
> available documentation if I want to understand the library. But
> that's backwards: I have not yet found any reason why I should want to
> understand this library.
>
> When I encounter a new Boost library, my first question is always: how
> would I use this? What problems in my work would be addressed by this
> library? How could it make my life easier? The closest I've so far
> gotten with Join is a vague sense that I could use it to implement
> Futures, if I didn't already have such a thing.
>
> Were I not already somewhat curious about ways to organize async code,
> I wouldn't have read even this much of the documentation.
>
> What's the point of this rant?
>
> - I'd like to request a bit more introductory material. If one of the
> References includes an excellent discussion of why I should be
> interested, please promote the link (a specific section link) to the
> Introduction. Otherwise, please add some new prose with specific
> examples.
> - I'm aware that this may be clearer to others than it is to me. If
> you understand the proposed Join library and feel its documentation is
> sufficient, now would be a good time to say so.
>
> My assumption at the moment is that if the Join library were accepted
> as-is into Boost, few people would even attempt to understand it
> because, like me, they would have a tough time figuring out how it
> might apply to their problem space.

This is actually one of main problem/issues I have with a lot of the 
boost libraries. I know there's some awfully good stuff in there, but 
with several of the libraries, the documentation does a poor job of 
telling me what it's actually all for. It must be interesting. People 
must be using it. But, why the heck would I ever want it?

I admit some of that is my problem. Or at least perhaps, I haven't come 
across any situations that would warrant one these libraries. The 
problem, though, is that if I did come across that situation, I wouldn't 
necessarily know that there is a boost library solves that problem.

-- Bill --