Re: Hi

Thanks for the info and advice.  Some follow-up is below; I have
edited out some of the original message.

On Mon, Aug 05, 2002 at 03:34:22AM -0400, Stephan Richter wrote:
> 
> > In contrast, my attention has been on managing bounce data for mails I
> > send out, and integrating mailman with ZODB both for messages and list
> > subscribers.
> 
> This is much harder than you might think, since you would have to rewrite 
> larger sections of Mailman itself or at least monkey patch some methods so 
> that they can reroute to Zope.

I think I'm probably going to extract just the bounce identification
code, which has a fairly narrow interface.  After having looked at the
mailman code, I can see going beyond that is getting into deep water.

> 
> > Another potential issue I see is that effective integration may
> > actually require ZEO, since both a running Zope and mailman's
> > responses to email may want to hit the database.
> 
> No, the ZODB transaction mechanism will handle that. However, you may need to 
> write an EventChannel for Zope 2. Look at Kapil's work.
> 

Where is Kapil's work?  I take it from what you're saying that the
mail processor could talk to the running Zope, rather than trying to
talk directly to ZODB?  I thought it was basic to the ZODB
architecture that two processes can't access the db at the same time.

An additional concern for me is that Zope may not be running all the
time.

> Overall, I think your project is fairly ambitous. If you do not need these 
> features directly for a project, I would suggest you help developing Zope 3 
> and talk with Barry a little bit about Mailman 3, since it will use many 
> parts of the Zope 3 architecture and will be therefore easy to integrate with 
> Zope 3.
> 
> Regards,
> Stephan

Developing Zope 3, or developing for Zope 3?  I didn't realize it was
so far along.

By the way (for anyone) are there any licensing issues with sticking
Mailman chunks into a Zope product?  At one point I had the impression
that the Zope and GPL licenses didn't quite fit together--though, of
course, products are in sort of a grey area to begin with (that is,
reasoning by analogy to the C compiler, it's fuzzy whether a product
is like a program written in C or an extension to the compiler).