Boost logo

Boost :

From: Jon Kalb (jonkalb_at_[hidden])
Date: 2001-11-12 19:43:52


It seems to me that you are advocating keeping the libraries from
becoming interdependent so that (new) users don't have to climb too
steep an initial learning curve--having to learn a large number of
classes to use one or two.

But it also seems to me that in the example you site, the
interdependency is on the implementation side, not on the interface. In
other words, I don't have to learn how to use Boost.Assert in order to
use a library that depends on it. There is certainly a build dependency,
but there is no requirement that I understand all the classes used to
implement a library in order to use it.

I also have concerns about dependencies, but I would rather have an
Assert system consistently used than to have each library roll its own
or *shudder* not use asserts.

--
Jon Kalb
> -----Original Message-----
> From: Thomas Matelich [mailto:toms-mailing-lists_at_[hidden]] 
> Sent: Monday, November 12, 2001 12:09 PM
> To: boost_at_[hidden]
> Subject: [boost] Interdependence of Boost libraries
> 
> 
> This is spawned by discussion about Boost.Assert, but I've 
> been concerned for a little while. Boost.Assert seems to be a 
> well designed system and I like it in general, but I'm 
> concerned about introducing it.  The point I'm going to try 
> to address is keeping Boost accessible to newcomers.
> 
> The best counter-example that I can think of is when I first 
> started to use ACE.  To use one class, you need to read up on 
> something resembling 500 other classes (perhaps a slight 
> exaggeration).  Obviously there is a purpose, and after 
> getting into it, I like ACE and appreciate its design.  On 
> the other hand, the standard library is pretty cool in the 
> other direction.  I can learn about iostreams and complex by 
> themselves, I can learn about vector without list, 
> algorithms, or auto_ptr, etc.  Of course as you learn about 
> more of the STL, you see how things draw together with one vision.
> 
> I realize I could use Boost libraries without caring about 
> Assert, but not quite.  In summary, I have no solution and 
> there is not a big problem, yet.  We've been working on 
> getting the website more navigable/accessible, let's try to 
> make sure the collection stays that way too.
> 
> --
> Thomas O Matelich
> Senior Software Designer
> Zetec, Inc.
> matelich_at_[hidden]
> tmatelich_at_[hidden]
> 

Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk