[Zope3-Users] So, there is going to be no Zope4... how to handle maintenance a couple of years from now?

[Milind Khadilkar]

Hi,

Seems there will be no Zope4. Puts a fond hope to rest.

I have one large Zope2 (Zope 2.6) project, one large Zope3 (Zope 3.4) project, two medium sized Grok projects, one GAE project, allof them need to be maintained beyond two years.

I would like, if possible, to redevelop them using ONE single framework.

While some of the original developers are available, most work would need to be done through people who need to be trained from the ground up.

My first choice (mainly because I have a complex Zope3 project to redevelop) would be Bluebream, even if it means using ZCML and programming in ZCA-shackles. But I am speaking without any real experience of it. (Negative press ensured that I did not go for it...)

Any suggestions?

Thanks and regards

Milind Khadilkar

[Thierry Florac]

Hi,

Le Sun, 7 Sep 2014 09:26:43 +0530,
Milind Khadilkar <zedo...@gmail.com> a écrit:

If you are considering a complete re-development, maybe you should have
a look at Pyramid.
It allows you to reuse a lot of Zope related technologies, including
ZODB, traversing and even ZCA ! ;-)

I use and develop ZTFY which is based on Bluebream (with updated
packages based on ZTK) and which is used in many of my own projects
with great success, but which is probably condamned to death in a
relatively near future...

Regards,
Thierry

P.S.: by the way, I don't really understand where is the real problem
with ZCA (which is so powerful, and not so complicated!) and with ZCML
(which is nothing in XML files terms compared to any "serious" Java
server development environment...)
_______________________________________________
Zope3-users mailing list
Zope3...@zope.org
https://mail.zope.org/mailman/listinfo/zope3-users

[Milind Khadilkar]

Thanks, Thierry.

I think the real problem with ZCA and ZCML is "bad press". We did not face any ZCML problem when we worked on the Zope3 projects. It was later, thankfully, that we came across the negative opinion. ZCA did require a mindset change for some, but more often than not it has helped them in their future work on other platforms too.

But that was 10 years ago... Don't know about now.

Regards,

Milind

[Thierry Florac]

Hi,

I think ZCA doesn't do "so much more" than implement a few object design
patterns and provide a very flexible components registry framework.
I'm working mainly on developing new framework-like packages and I just
can't work without it except by re-inventing it!

Many users blame ZCML just because it's XML !
But it's just configuration, and separating configuration from
Python code is one of the greatest benefits of ZCML, mainly while
working on huge projects. The XML format is just a detail and is
probably not so worse than many INI-like or text based configuration
format; and at least it's a de-facto standard.

Regards,
Thierry

Le Sun, 7 Sep 2014 14:14:14 +0530,

[Christopher Lozinski]

On 9/7/14, 6:56 AM, Milind Khadilkar wrote:

Hi,

I have one large Zope2 (Zope 2.6) project, one large Zope3 (Zope 3.4) project, two medium sized Grok projects, one GAE project, allof them need to... be redeveloped using ONE single framework.

Any suggestions?

In general the following guidelines apply.  If it is heavily relational database/ URL dispatch dependent, go with Pyramid. If it is pure file system python with traversal, go with the simpler Grok.  if you are doing a lot of TTW work, acquisition and Zope 2 Security interface, then your best option is Zopache on top of Grok.  Of course real world situations are not so clear.  An informed choice would take the following factors into account. 

How many lines of code/classes are there in each application?  What security model are you using in each case? View based or traversal based?  Do you need the Zope 2 user interface for configuring security? How much code is done Through The Web, (TTW), and how much on the files system.   Are you using Acquisition?  For TTW code, which classes are you using?   DTML? Zclasses?  How is the GAE application architected?  Is it even Zope compatible, or is it a relational database and URL dispatch kind of application?

Given the answers to those questions, it should be quite clear what your best strategy will be. 

My first choice (mainly because I have a complex Zope3 project to redevelop) would be Bluebream, even if it means using ZCML 

If you are going to start with BlueBream, better to start with the ZTFY.org.  and wiki.ztfy,org.  I wrote most of that wiki.  It is a much more modern and up to date than the most recent bluebream distribution.

Better yet, hire Thierry the author of ZTFY.  Let him work from Paris.  He indicated that his current project is coming to an end.   My highest respect for that man and all he has accomplished.  His tech support was brilliant.    I kid you not, he would be at lest 10 times more productive than any Indian developer you might hire and try to train. It would take a few years to train someone to replace him.    The man thinks in Interfaces.  

So ZTFY is better than BlueBream.  But using Grok is better than straight ZTFY.  Why?

While you and Thierry have had good experience with ZCML, let me assure you I have tried bluebream, ZTFY, and Grok, and Grok is way way easier.  Ask anyone who has done both approaches.  ChrisM the author of Pyramid wrote an excellent analysis of the difficulty with ZCML in the top part of defense of Pyramid
http://docs.pylonsproject.org/docs/pyramid/en/latest/designdefense.html

Do read it.  Particularly in your case, where you talk about hiring new developers and training them.  Grok hugely simplifies application configuration.  It feels just like writing regular python code.   And reducing the conceptual burden on new developers is a huge issue in this Zope world.  Of course if you go with a senior developer like Thierry, then starting with ZTFY is acceptable.   Although I would argue that even seasoned zope developers would be more productive in Grok. 

Now what about your Zope 2 application?

You said your largest application is the Zope 2 application.  Are you using acquisition?  Lots of TTW stuff?  Clicking on tables to define security.  Then Zopache.com with Grok is the tool for you.    Zopache is the cultural inheritor of those software approaches.

Anyhow I was quite serious about my questions at the top of this email.  A bit more information about your applications and how they are architected would help enormously in figuring out what you should be doing.
 
 
Hope that helps. 
Chris

















 

[Jim Fulton]

On Sun, Sep 7, 2014 at 5:37 AM, Thierry Florac <tfl...@ulthar.net> wrote:
>
> Hi,
>

> I think ZCA doesn't do "so much more" than implement a few object design
> patterns and provide a very flexible components registry framework.
> I'm working mainly on developing new framework-like packages and I just
> can't work without it except by re-inventing it!
>
> Many users blame ZCML just because it's XML !

As well they should! 0.5 ;)

> But it's just configuration, and separating configuration from
> Python code is one of the greatest benefits of ZCML, mainly while
> working on huge projects. The XML format is just a detail and is
> probably not so worse than many INI-like or text based configuration
> format; and at least it's a de-facto standard.

When I started working on zcml, it used a very light-weight but custom
syntax. I was talked in to using XML by someone who later complained
about it. :) I regret not sticking with the lighter weight syntax.

Jim

P.S. ZCML uses a pretty lightweight style of XML, relying more on
attributes than on nested elements. Ironically/typically I was
criticized for this by XML purists.

--
Jim Fulton
http://www.linkedin.com/in/jimfulton

[Jim Fulton]

On Sun, Sep 7, 2014 at 4:44 AM, Milind Khadilkar <zedo...@gmail.com> wrote:
> Thanks, Thierry.
> I think the real problem with ZCA and ZCML is "bad press". We did not face
> any ZCML problem when we worked on the Zope3 projects. It was later,
> thankfully, that we came across the negative opinion. ZCA did require a
> mindset change for some, but more often than not it has helped them in their
> future work on other platforms too.
> But that was 10 years ago... Don't know about now.

I'm proud of the ZCA in many ways, but:

- ZCA was designed for problems that most people don't or shouldn't have,
which is making a complex application pluggable.

If your application is complex, that's a problem

http://legacy.python.org/dev/peps/pep-0020/

- In the Zope 3 project, we used the CA way more than we should
have. Initially, this was to prove that we could. Once we were
convinced that every (damn) thing could be pluggable we should
have stopped and simplified, using the ZCA only where needed.
Instead, we'd established a culture of crazy levels of indirection.

- Outrageous indirection in the base system made starting new projects
either super difficult, an exercise in cargo-cult-programming, or both.

I've come to the conclusion that any framework that requires or
encourages its users to use project-templates or project-setup
wizards isn't something I want to use.

I stopped using Zope 3 several years ago when I realized that the
weight of the framework wasn't justified by it's benefits, at least for
me.

I've decided that I'd rather use decoupled frameworks that, ideally,
are simple to learn and use individually. That's why I use bobo now,
http://bobo.digicool.com. A more conventional choice along the same
lines would be Flask, although I think bobo is simpler. (Of course, I'm
biased. :)

I still use the ZCA, especially zope.event, but in a wildly lighter-weight
fashion than I did in Zope 3.

Part of the reason I prefer simpler server frameworks today is that
Web applications are far more client centered today than they were
when I worked on Zope 2 and even Zope 3. Today, for applications
(as opposed to content *sites*), UI logic, including templating, mostly
happens on the client, and web servers are largely REST/RPC/Database
servers.

Jim

P.S. If you find the ZCA interesting, you should check out Scala and it's
implicits and type-level programming. It does many of the same
things as the ZCA at compile time. It's crazy beautiful and evil all
at the same time. :)

--
Jim Fulton
http://www.linkedin.com/in/jimfulton

[Milind Khadilkar]

Thanks, Jim, for your responses!

These thoughts from you deserve a larger readership than this thread provides.

I think I will restart thinking on this from scratch.

Thanks, again.

-Milind

[Milind Khadilkar]

Thanks, Chris

I agree with you on most counts!

I will add that in my limited interaction with it, I found grok "beautiful".

I need to rethink on this from scratch. Fortunately I have time till the end of the year to help ourselves decide.

I was not aware of Thierry's ZTFY. I will definitely look at it.... and take one more at Grok. I have just read about Zopache a few days ago.

-----

One of our largest applications is in  Zope3, and another is in Zope2. However, in both cases the domain modelling complexity and SVG based UIs (and the need to thwart screen-scrapers from getting at the raw data) outweighed the architectural issues. Yes, for the Zope2 project we did use TTW a lot, used DTML and ZPT in tandem, and acquisition too. The GAE project is straighforward, with Google's object database, ZPT... partly it was ported from Grok. But I will have to dig up the old documentation before I can answer your questions with confidence.

Thanks, again.

Regards,

Milind.