cxf-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Liu, Jervis" <j...@iona.com>
Subject RE: tooling and service model
Date Wed, 20 Sep 2006 07:01:10 GMT


> -----Original Message-----
> From: Dan Diephouse [mailto:dan@envoisolutions.com]
> Sent: Wednesday, September 20, 2006 1:00 PM
> To: cxf-dev@incubator.apache.org
> Subject: Re: tooling and service model
> 
> 
> James Mao wrote:
> 
> > Hi Dan,
> >
> >>
> >> I would like to start a set of parallel tool modules. Say create a 
> >> directory called tools2 and get things working there. Once we're 
> >> ready we can work on switching the whole build to the new tools.
> >
> > The tools2 will include all the functions in cxf tools?
> 
> And more :-)
> 
> > and how long will it take? 
> 
> Depends on how fast we can type :-)
> 
> > and when we say switching you mean include porting all the 
> unit tests 
> > and system tests to the tools2? and also make demos work properly.
> > So, your goal is to replace cxf tools with tools2?
> 
> Yes, yes, yes.
> 
> > or let user choose which one they want to use?
> 
> No.
> 
> > But i really have a question why we need to create another 
> tools when 
> > the current tools works perfectly fine?
> 
> Because they don't work fine. I believe we had this 
> discussion before... 
> Please see the previous threads on tooling.
> 
I do remember we had some discussions before regarding CXF tooling, but as far as I can recall,
I do not think we are anywhere close to reach a conclusion that CXF tools need to be totally
rewritten. 


> If you want to try to refactor the current tools in place, I 
> suppose we 
> could try that. But its a pretty drastic change, so I don't 
> really see a 
> step-wise process to do it. I think it would be better to 
> just write a 
> second set of tools, then when they're ready switch 
> everythign over and 
> delete the old tools.
> 


Refactoring current tooling or writing a brand-new one are both ok as the ultimate goal is
to have a better tooling. But it would appear to be too early to tore down everything we already
have and invent the wheel from scratch before we can answer questions below:

1. Identify exact problems in our current tooling, and what problems can be resolved through
refactoring and what problems can not be resolved elegantly as it is due to core architecture.


I just went through our mailing list and found following tooling relevant discussions, do
let me know if I am missing anything.

a. CXF tools need to support WSDL2.0
b. CXF tools need to support SOAP1.2
c. We need to reuse Service Model in tools, this will bring us WSDL2.0 and SOPA1.2. There
were some discussions around how to do this. One proposed approach is moving Service Model
as a top level module, one is using Service Model as a plug-in. To me, they are implementation
details and we shall be able to figure out which way is more appropriate when we are doing
this work. 
d. Support multiple databindings and multiple front-end profiles generations. With the code
merged from XFire code base, the current CXF tools should have this capability in place already.


Based on this observation, I feel a refactoring is a more practical approach, doing (a)(b)(c)
should be a one week work. At the same time, I have to say some interfaces in our current
tools are a bit confusing, for example, the ToolsContainer interface. This makes new developers
hard to following the tooling architecture in order to do their contributions. I suggest we
do another refactoring on this after getting service Model done. But overall, I do not see
any critical problems that can not resolved through refactorings based on the information
I have right now. 

2. If we are to write a new tool from scratch, what are the feature list we have in mind,
and how long do we expect to reach this feature list. 

To be honest, I do not really think writing a new tool and make it has all features we already
have is that simple. For example, current wsdl2java give us following options:

Usage : wsdl2java -p <[wsdl namespace =]Package Name>* -b <binding-name>* -d <output-directory>
-compile -classdir <comp
ile-classes-directory> -impl -server -client -all -ant -nexclude <schema namespace [=
java packagename]>* -exsh <enable
extended soap header message binding (true, false)> -dns <Default value is true>
-dex <Default value is true> -validate
-h -v -verbose -quiet <wsdlurl>

In my experience, a good tool can not turn out in weeks or even in months(presume it already
has a good architecture core), it is a time consuming work, it needs daily and sometimes tedious
work to add features, fix bugs and tune around based on users/customers' input and real use
cases.  Looking at the type test, interop test and the set of system tests (which are based
various bugs discovered by ourselves and customers/users), you will probably feel our current
tools are gradually reaching a mature status. I am pretty sure writing sth from scratch can
always bring better architecture as we can learn a lot from past, but what is cost? How long
do we expect the new tool can reach a same mature status?



> >>
> >> I'm in the midst of refactoring this now. I probably won't get the 
> >> code in until tomorrow though. I'll try to explain once I get it 
> >> finished :-)
> >>
> > Cool, just make sure all the system tests and demos working 
> properly.
> 
> I don't usually test demos when I do commits. That would really slow 
> down the development process. If there is something in the demos that 
> isn't covered in the tests, you should add it.
> 
> - Dan
> 
> -- 
> Dan Diephouse
> (616) 971-2053
> Envoi Solutions LLC
> http://netzooid.com
> 
> 
Mime
View raw message