Re: [boost] Re: Any interest in adding unicode support to boost?

From: Peter Dimov (pdimov_at_[hidden])
Date: 2004-10-20 14:51:54

Stefan Seefeld wrote:
> another argument is legacy / compatibility. I'v written an XML library
> in C++ that essentially wraps a C library (libxml2 for the insiders),
> and so I'd like to avoid unnecessary string copies when passing data
> around.
> While the inside uses utf-8, I want to make the wrapper API adaptive
> to user-side unicode APIs. Thus the unicode type has become a template
> parameter, and conversion is being done by a trait class.

If the inside uses UTF-8, you should just accept UTF-8. The standard way to 
avoid copies is to provide a function/constructor that takes an iterator 
range. Users can now pass anything and aren't forced to exclusively choose 
one encoding, which becomes a part of the component type and can never be 
changed.

To paraphrase, unnecessary parameterization is the root of most evil 
nowadays. Parameters are a liability, not an asset.