Re: [boost] Re: Re: Adding Loki to Boost (reprise)

From: Mark Rodgers (mark.rodgers_at_[hidden])
Date: 2002-04-12 14:51:41

----- Original Message ----- 
From: "Powell, Gary" <powellg_at_[hidden]>
> First:
>   Did I miss something, or is mpl up for review? Because there are a lot
> of mpl reviews being posted.
> 
> Second:
>   Did I miss that Loki is up for review? Because its being compared to a
> library which hasn't been released. 

Hmm.  Did I miss that the only discussions we're allowed to have is to
review libraries?  Andrei made some pretty provocative statements that
needed to be challenged, so I don't see the harm in discussing them.
Hopefully we can convince Andrei to open his mind a little.

> Thirdly:
>   Is there some hard rule that boost can't have two typelists? Why do we
> have to have one mega all encompassing solution to all problems? Isn't
> it possible to have a simpler typelist and a complex one? Screws and
> nails, each has a use.

I don't think there is any need for two pretty much identical typelists
and I fervently hope that we don't end up with two.

The question is whether we need two sets of algorithms with pretty much
identical interfaces but different implementations.  I fail to see the
benefit in that.  I almost never look at the implementation of STL 
algortithms, and I don't particularly care how MPL algorithms are 
implemented as long as I can use them easily, particularly with MSVC*.

Can you seriously imagine anyone looking at their local STL implementation 
of std::distance and saying "Wow, all this dispatching on iterator tags 
is *so* complicated, I can't use it.  I'll have to create my own simple 
implementation I can understand and use that instead."  And is anyone 
seriously claiming that std::sort is only useful because the
implementation is simple?  I don't think so.

Andrei, your opposition to MPL really smacks of NIH syndrome.  I really 
do hope that you can bring yourself to use someone else's code and 
contribute parts of Loki built on top of MPL.  What harm could it do?

Mark

* MPL does support MSVC, doesn't it?  If so, comparisons of MPL count_if
with a version that relies on partial specialisation is _completely_
invalid anyway.