[XML-SIG] Re: Maki: great!

Nicolas Chauvat Nicolas.Chauvat@logilab.fr
Thu, 21 Feb 2002 13:23:06 +0100 (CET)

> Zope:
> I tried it a year or two ago. I abandoned it for these reasons:
> 1) The existing documentation was scarce and not very clear. I had to figure
> out how to use, configure and extend the system by trial and error. I had a
> hard time trying to understand the whole architecture of Zope and its "nuts
> and bolts". For sure, the current situation is much better.

Yes, Zope documentation is getting better, but I agree with you that it is
really not good to have to spend so much time to get the "Zope Zen". I
wrote Products for Zope and I'm not Zen about it...

> 2) You have to use its UI to create a site, configure and manage it. I was
> unable to find a way for creating pages, configuring the system and manage
> it with external programs/scripts or with external tools. In my production
> environment, it is not very realistic that every single operation on the
> system has to be performed by hand.

I agree with this too.

> 3) The OODB behind Zope is quite an obscure object and I found impossible to
> access it in any other way than using the supplied Zope GUI. That is quite
> limitating. Think to a massive "import" from an external data source.

Same impression here. But ZODB was recently spun-off as a stand-alone
project and that may improve things.

> 4) The Zope architecture is quite monolithic. I was unable to understand how
> to extend it and how to create task-specific modules, probably and partially
> because of the scarce documentation. I'm sure the situation is much better
> now (You mentioned "components"...).

Zope components (called Products in Zope talk) are a reality. They are far
too difficult to build, IMHO, but I don't now of other Pyhon-based web
framework that provide such reusability. I haven't used that many
different ones, tho.

As a conclusion, I think Zope needs much improvement indeed, but it's
there and it works. Is it worth it to start completely from scratch ?

> 4SuiteServer
> 1) 4SuiteServer is a commercial application. When possible, I (we) avoid to
> use commercial application, even if the cost is very, very low. The reason
> of this (arguable) behavior is that is becoming a nightmare to manage the
> huge amount of license documents/criteria and the continuos stream of
> licence expiration/payment events in our (already chaotic) production
> environment. For what regards me, I abandoned M$ for the same reasons a long
> time ago.

Beep [wrong answer]. 4SuiteServer is open source software ! Nothing there
can be compared to M$ !!!
> What I really need, is an open source, modular, easily configurable, easily
> extendable, pre-built portal. Something I can install in a couple of hour,
> prepare it for the customer in a couple of days and manage from script and
> external tools at my wish. Something that makes easy to create task-specific
> modules and that can accept new modules in a easy way. That is exactly what
> ezPublih gives me in the HTML field. That is what "Cocoon portal" promises
> in the XML arena. As long as I can see, nothing like that is available in
> the Python world, at the moment.

"Build it and they will come." :-)

Nicolas Chauvat

http://www.logilab.com - "Mais oł est donc Ornicar ?" - LOGILAB, Paris (France)