[Zope-dev] Re: [Zope 2.12] Eggification of Zope 2 - pros and cons?

[Chris McDonough]

Chris McDonough wrote:
> FTR, there are things in Zope 2 (like Missing and Record IIRC) that 
> depend on ExtensionClass (or Acquisition) headers, and there is no way 
> to tell setuptools to depend on an external package to provide 
> compile-time headers.  We could fake it by including externals in Zope2 
> svn for these headers, but then there's version dependency "hidden" in 
> these externals that will be violated if the EC and Acqusition eggs 
> change in any given setup.  There's certainly no hue and cry from the 
> masses I've heard that EC and Acquisition be usable outside Zope 2.    
> For this reason, I'm not entirely sure it makes sense to break 
> Acquisition and EC out of a larger Zope 2 package.  Likewise for 
> DateTime (given that there's already a Python datetime).  I suspect it 
> would be decomposition for the sake of decomposition, which is not very 
> compelling.
> 
> There are is a similar problem between things in Zope2 and ZODB, but 
> ZODB does have a life outside Zope2, so I think it does make sense for a 
> Zope2 depend on an external egg for ZODB packages.

Likewise for Medusa.  This is already packaged as an egg, we just need to delete 
it from ZServer.

- C