But that isn't what he thinks he wants, :-)
Allow me to speak for what i want.
I am glad to hear that people think that there is a role for TTW editing. I am glad that there is a tool in Zope 3 for TTW editing. http://agendaless.com/Members/tseaver/software/userschema/ I am glad that there is a tool in Plone for TTW editing. ATSchemaEditorNG I wish there was a supported tool in Zope 2 for TTW Editing. Even better would be linux GUI development environment. It could also run on Windows. But that is just a dream.
Sure there are python editors, that minimize the number of mistakes I can make, but it is better to have direct Zope editors, where the only things I can do are the ones that Zope allows me to do. I took a quick look at the CPS documentation in french, my french is not that good, it still looked like lots of opportunity for me to mistype python code. Maybe I was looking at the wrong page.
And I would like these schema generators to understand not only specific python objects, but also the network of python business objects. I think that in Zope 3 a little bit of that is in there. A choice list for a field can be generated from some other list of objects. But I would love that network model of objects to be more deeply embedded into these frameworks. Specifically I am less interested in the interfaces, and more interested that a semiconductor factory can contain etchers, diffusion equipment etc. I am less concerned about the interfaces, and more concerned about the business relationships between objects. I am not quite sure how I want to specify that in software.
I would like a security model based on business rules. I loved Zope 2, but I now understand and need more. I want to build a security model based on my business model. For example, Clients in a company who have the accounting privilege can see but not edit their companies invoices, and only their companies invoices. Recruiters can generate invoices, and see the account status for their clients, but not for other companies. Supervisors can see all the invoices pertaining to them, and see the financial reports for the recruiters they supervise. That is way different from the Zope 2 security model. Sure it can be implemented in Zope 2, but it is tricky, complex, prone to errors, hard to debug, and hard to test. I think Zope 3 comes closer to allowing me to build my own security model.
Of course I would love such a security model in Zope 2, but I am not waiting for it. Maybe I could do it with 5.
I would love my existing ZClasses to continue to work.
There is another big issue. I like to minimize the number of lines of code in my application framework. This just minimizes my complexity. My pet theory, for which I have no proof, is that there is just too much stuff in the stack python/zope/cmf/archetypes/plone for it to be simple. For my next app the only thing I have approved so far is python and ZODB. The Zope 3 Schema component also looks interesting. I need to look at it more closely. Maybe Zope 3 Security will make it in. Keep the platform simple is my plan.
I am just not willing to grab the whole Zope 3 bundle of stuff, and then add Grok on top of that to make it simpler. The decisions I make now, I have to live with for a long time. The need for grok
is a signal to me that the Zope 3 abstraction is not the abstraction I am looking for.
A much simpler solution for me is to build a business model on top of the ZODB, and then just access the few user interfaces from Zope 2.
If you want to read more about what this zope customer is currently thinking you can read more here.
I will take a moment to mention a python gui editor for a tree of objects. It is GPL. It would be interesting to hook this up to the ZODB.
All your feedback, while difficult, has been very helpful in refining my thinking.
Zope-Dev maillist - Zope-Dev@zope.org
** No cross posts or HTML encoding! **
(Related lists -