Philipp von Weitershausen wrote:
Jim Fulton wrote:
Here's what I'd like to see:
1. Separate server definition from application definition. They are way too interwoven now. I'd like to see us move to Paste Deploy (or something based on the framework it defines).

Is this the seperation of "what objects are being served" from "what protocols are being served" that I mentioned in my other reply?

Someone. More it's a separation of the server definition from the application definition.


I suppose, you would like to do the server definition thru something like paste.deploy (allowing us to let the assembly currently done by the two main programs be done generically).


What would the application definition be? Still ZConfig?

For the use case of a ZODB-less application, no.

In the short term, for me it will be less work to use ZConfig
with the schema defined in  Eventually,
I'd prefer to move to something simpler and more flexible.

Paste Deployment has a simple model for passing application
configuration data. I'd explore that first. It may be too simple.
It's hard to tell from the fairly thin documentation. I haven't
actually used Paste Deployment myself. :)  It does look like
something we can build on.

(Note that in my original reply to ChrisW, I tried to be pragmatic. Of course, if ChrisW has the resources to go into this refactoring, by all means, let's do it).

The trouble is that out existing main programs are the result of
layered pragmatic decisions to build on what we already have,
resulting in something that is a bit of a mess.  I don't mean to
lessen the contributions of those who got us where we are, but
I think it's time to do something different.


In order to implement a simpler security mechanism for grok, we had to re-implement every single publication (though of course subclassing was still possible, but we had to create a new variant of every publication class). That shouldn't be necessary.

WRT ZConfig. I find myself fighting ZConfig a lot. If I were to write a new main program, I wouldn't use it at all. In fact, I would probably use a much simpler model like that used by Paste Deploy. I really don't want to have a ZConfig argument though. I got out voted on that last year. If you do want to use ZConfig, you could simply use the application schema defined in rather than the ones in the server packages.

I think the ZConfig argument was largely due to misunderstandings. I would be surprised if people really cared whether to Zope used ConfigParser or ZConfig (except Fred, perhaps ;))

That's not the impression that I have.  I've resigned myself
to having the existing main programs continue to use ZConfig,
in part because I'm fairly sure they will eventually be irrelevant
to me.


Jim Fulton           mailto:[EMAIL PROTECTED]       Python Powered!
CTO                  (540) 361-1714  
Zope Corporation
Zope3-dev mailing list

Reply via email to