Re: [boost] date_time naming

From: Jeff Garland (jeff_at_[hidden])
Date: 2003-09-03 09:31:55

On Wed, 3 Sep 2003  Paul Bristow wrote

> Although I an growing to like date_time, I have to agree that some names are
> less than ideal.  I found kday less than intuitive.  

This name actually comes from Calendrical Calculations.  But I'm not stuck on
it if you you have other suggestions.

> Documentation of the labyrinthine (with good reason) structure is also 
> weak (or even wrong?)

I'll be the first to acknowledge that the documentation needs to be improved,
but honestly I'm much more focused on improving other elements of the library
at the moment.  To me the parts of the library that are basically undocumented
are implementation details subject to change at any time. This makes it
difficult for others to extend the library and I acknowledge that, but
normally what happens is I either help or some users can actually figure it
out.  At the end of the day, it is an unfortunate but true that there are a
finite number of hours in a day, I have to work for a living, and also have a
life that does not involve staring at a computer screen.  As a result I
suspect there will always be a gap between what I and users would like to see
here...

>It seems to me that these observations at this stage highlight a weakness 
>of the current review process.  Until code gets Boost acceptance status, 
> too few are prepared to really use it in anger on real projects, and only

This is true, many users won't use it until it is part of boost.  That said,
you might be suprised at the number of early adopters prior to being part of
boost.
 
> then do lots of 'issues' start to surface.  But by then, changes cause 

This is a natural part of software development.  As more and new users bring
new perspectives new issues surface.  Implementations get rewritten as new
knowledge is available.

> grief to existing users, so there is a reluctance to 'improve' things 
>like naming.

I have no problem with deprecating existing naming.  Other libraries (eg:
function) have done this as well.

>Do we need a 'still may be subject to significant change' status to
>distinguish from a 'pretty much tried and tested' status?

No I don't think this is possible to measure.  I think the regression tests
are the real measure of this.  

Jeff