Hello,

I'm new to Hibernate and I am begining the design of a client/server swt app.
Something like that
[swt client] <--->[db & appServer]<--->[other swt clients]

Different from most of the discussions I found in this forum, I have total control to the application architectural design.

The main concern is a list (swt table) the user uses to search/select an item.
I think the view code will request rows on demand when they become visible on the screen. The view code will request row data to a "data source".
The data source will retrieve the requested data from a database table that has more then 200.000 itens (the user can set filters to narrow the search but still can result in more than 1000 itens).


The question is how this data source should work?

Approaches I am considering now:
1 - Use Hibernate directly on the client. I will open a scrollable resultset and scroll to the requested row as needed.
2 - My data source will connect to SessionBeans on the appServer, those SessionBeans use Hibernate.
3 - Create a RMI server. The server objects will do the same as the SessionBeans (except for CMT - not used in this case, but useful in other parts of the application).

I think 3 is not a real option to me unless it has many advantages over the other two.

What are the advantages and disavantages of each approach? And are there other approches I am not considering?

Does make sense to retrive rows in chuncks from the server (the SessionBean reads a page at a time and return an ArrayList of rows to the swt client)?

Any thoughts or suggestions will be of great value to me.

Clovis.

using rmi may lock you into a propriety protocol, also be aware that at least on my company, sending an object from a java client to a server took 75 seconds!! (we still don't know why we think it might be somethign we DNS).

thanks for the reply.

75 seconds is prohibitive for any application you should have.
Do you meanit tooks 75 seconds to get initial reference to the object? and what about subsequent method calls? Clients are local or remote from the server?

I used RMI in the past and notice no delay on a Local Area Network.

I forgot to mention that my application will run on a LAN (100MB) or WLAN (shared 54MB).

for the first method call it took 75 seconds, however subsquent calls (with the same object) took fine).

client are remote from the server, it also changed according to the platform, RMI from one windows workstation to another to 20 seconds, from windows to a unix server took 75 seconds.

we also used fixed ip, to maybe avoid DNS issues.