0==========
原文(fancyh于2001/04/30 17:12粘贴) 
How to design EJB patterns?Xingzeng,typet1998 can you tell me? 
--------------------------------------------------------------------------------
We will start a new project recently. 
We decide to use weblogc,Apache,Lunix,win2000,sql server. 
We have some basic knowledge and skill of OOA and rose. 
We have basic J2EE knowledge such as ejb,jsp,servlet,java bean. 
And we have also test those technic in simple ways. 
I also read some tech doc from web on the thinking of ejb pattern. 

So I want some more useful ejb patterns from successful project 
to avoid fault in devlop. 
I think the architectural of the full software is important. 

Can you give me some advices or resources about it? 

Tks!! 

1==========
原文(fancyh于2001/04/30 17:13粘贴) 
xingzeng ,look here 
--------------------------------------------------------------------------------

2==========
原文(fancyh于2001/04/30 17:14粘贴) 
typet1998,look here 
--------------------------------------------------------------------------------

3==========
原文(fancyh于2001/04/30 17:15粘贴) 
Skilled ejb designers,OOA experts,look here 
--------------------------------------------------------------------------------

4==========
原文(xingzeng于2001/04/30 17:59粘贴) 
The J2EE official sample could be one of the best learning source. 
--------------------------------------------------------------------------------
Using EJB could be the easiest way to build modern application. The blueprint sample Java Petstore can help anyone to understand the way to design well structured and well performed application. 

One major problem for EJB is to use or not to use container managed entity beans. This really depends the situation of your project. Do you really want to have your project done within one month and first ignore performance? Then CM bean could be the solution. My article in "Database proramming with EJB" published in "Enterprise Java developer" could help you solve many tricks in CM entity beans. Most case I still use Bean managed entity beans. A little bit more work, but more flexibilities. If you use SQL Server, the best performace of SQL server can only be achieved by using integer primary keys. So using identity for primary key is the natural selection. So working according petstore sample to build the data acess object is a really good idea to maintain the future portability. 

Probably you will use Win2K as the server, in that case you may need more time to investigate the Weblogic. Sometimes I get problem to deploy the EJB in Weblogic or Websphere with Win2K (NT has less problems). With Unix you do not have this kind of deployment problems. Some problems are caused by the port configuration.