Re: [glob2-devel] Refactoring

[Stéphane Magnenat]

> I remember Steph saying that: rewriting glob2 core code isn't worth
> the effort.  On the other hand the effort to rewrite new code is
> reasonable !?!  I don't get it.

I've said this, but our current cleaning of unit allocation and unit state 
machine turns indeed in a rewrite. Problem is that it takes long time.

The question about dependency is that one: if rewriting a feature takes 10 
minutes, as it does for the printf stuff, and adding a dep will take one hour 
to each porter, then rewriting is easier. If we want to add 20 of those 
features, than dependency is better. Furthermore, the more dependency the 
less people compile glob2 easily, especially on OS X (where we do not have 
any package right now) and win32.

Now for the precise issue of printf, at the end, I do not care either to use 
boost or to use a small rewrite. I just suggested how it could be rewritten 
to address some complaints about dependencies.

Perhaps some of you will complain about the lack of strong leadership (as my 
student do, it's frightening how much they ask to be coerced into working, 
very sad), but I think that everyone should find its own path in the art of 
computer programming (and, as speaking about it, this stupid assembler of 
60ies Knuth is using in his books is very unfriendly).

Steph