Julian13 wrote:
floater, very brilliant perspective. I'm looking forward to see your writing.

which Jakarta Struct book you mean? the John Wiley & Sons or the Oreilly one? i would like to have a look of it.

BTW, the idea of LogicSheet and XSP in Cocoon do release her from one-way pipeline. However, it seem not very handy to be used. May be more time I should invest on it.

floater wrote:
I think it's in the Oreilly book.
... REMOVED TO SAVE SPACE ...
contribute to it. When the time is ready, we could go forward.

Julian13 wrote:
I am more on utilizing the framework instead of developing mine in this stage (though it may be fun and satisfactory ^.^).

I am using a set of tools generating my domain class, data-access class, HTML, JSP and servlet for database centric web application. It does work well in term of user experience since I have added different JavaScript and CSS stuff for interactive user interface (to avoid massive reloading).

Neverless, I am facing difficulty to enrich my backend side like server side validation and connection pool, etc. I believe these can be solved by some framework Thus, I am more focus on the aspects like customizable workflow and business logic. I expect the framework allow me customize the application workflow and business logic on-the-fly with siimple script plus utilizing most of my domain classes.

As you said, XML/XSLT based templates might be the ultimate solution. However, the XML/XSLT itself might be too generic as a framework [for me]. If it can support some open standard XML format world be better. For example, if a framework support XMLSchema to dynamic create HTML frontend and database backend or a framework support BPML (http://www.bpmi.org/) for dynamic processing workflow, etc. This really boost up the speed for development and help developer more focus on the business logic instead of the technical issue.

So far, I think Cocoon is a excellent foundation to what I want. Of course, I will try to study what you have quoted for our reference too. Thank you very much. ^.^

coder wrote:
struts不错，不过我不喜欢jsp和taglib，所以，我找了一个strutscx，正在看。
http://it.cappuccinonet.com/strutscx/

特征：
综合运用struts、xml、xsl，No JSP

其它struts+xml的site
http://www.oroad.com/opencode/stxx/index.html
http://www.orbeon.com/oxf/

coder wrote:
http://it.cappuccinonet.com/strutscx/
http://www.oroad.com/opencode/stxx/index.html
http://www.orbeon.com/oxf/

coder wrote:
不过我不喜欢jsp和taglib

Julian13 wrote:
I don't like TagLib too but find JSP okay for prototyping. Both are good for initial prototypying and presentation purpose. However, the application become hard to maintain when the size growth. The client and server side mess up like a chaos.

I am recently doing a intranet purchasing system for the employee placing order. Since these ordering works are manual intensive, the client require a more interactive and user friendly GUI, for example, text box pop up with available value filtering, auto-filling related field (like selecting template order then fill up the new order), etc. I believe these requirements are typical for client-server application but not so common for the stateless web environment. So, I expected some of the web framework or application has addressed these issue (bad that I'm wrong). Do you think these requirements should be addressed by a framework?

For my initial solution, I have a set of JavaScripts and JSPs generated according to the database schema such that my HTML form can utilize them for auto data retrieve and form filling (They work closely with some auto-generated servlet, domain and DAO classes as I mention before). A prior solution about JavaScript-Applet has been done too. However, it has higher barrier than using JavaScript and JSP. My generated code work siimilar to the following links: http://www.aspfree.com/authors/Dario_Medele/masterdetail.ASP
http://ashleyit.com/rs/jsrs/select/php/select.php
Any comment for my solution?

Hopefully my work can be saved from a better framework in near future.

floater wrote:
I also have an inquiry about your code generating code. Suppose you finish a project like this, use a bunch of class to generate other classes.
How hard to modify the generating code to get the generated code if there is some change? I have no experience on this, but I am wondering how easy to break some code when you modify the generators because everything is tied up. And the debug is not easy, I imagine. Just curious. I always believe the rule: factor out the invariants and work on the variants.
But this kind of code-generates-code approach at the application level is kind of interesting(applications are a little bit unpredictable, nothing like from jsp to servlet rules).

Julian13 wrote:
You are right that it's easy to break the final code if anything goes wrong in the generator. however the same apply if a bug is fixed in the generator, it populate to all the final code too. ^.^ the generator is not any innovative application; or it's simply performing some dummy text appending job. however, it does saving my time from lots of typing. especially as my code has more pattern being used.

admitted that the debugging task is not trivial at the first thought. nevertheless, you can separate out the final code from the generator. For the final code, I'm using JUnit to test the business logic (basically the database CRUD + master-detail stuff). For the generator, I'm using some program to diff the generator code from my manual-typed code. such that I can focus on the business design once the generator code settle down.

the extreme case is I have not enough time fixing the generator. thus, i can manually change the generated code as no choice at all. of course, fixing the generator seems more productive in my case.