>> - I do not like to work with request and response directly, 
>> this is too low-level to build real applications. Seaside 
>> and SmallWiki both have nice concepts to avoid working with 
>> header-fields.
>
> what  is thekind of mechanism they use to handle with header fields?

Assigning a block to a link, button or input-field. When the page is 
submitted or the link is requested, that block will be evaluated and all 
the necessary arguments will be prepared and passed into that block.

>> - I want to debug/test my code, but Zope doesn't allow (out of the 
>> box).
> 
> false, it's a python program, and you can use the normal pdb, the 
> debugger that comes with every python installation.

Yeah, I know, but have you ever tried to do that? When developing a web-
application I usually don't want to go through the code of Zope itself, 
but through my code: it is almost impossible to debug the DHTML or the 
ZPT rendering!

> but a developer can have a good "web" productivity with zope, when he 
> knows the beast..

I doubt so ... 

- Something what I always missed in Zope is that Python is unable to 
tell if a particular file is still being used by some other component or 
not (like the senders of a message).  

- Furthermore do you really think that doing string-matching to read 
header fields of a request-object makes your code reusable. It is very 
hard to find bugs, when passing things around using that low-level HTTP 
protocol. I know there are also session-objects, etc. but this is not 
really a nice solution ...

Cheers,
Lukas

-- 
Lukas Renggli
http://renggli.freezope.org