[eclipse-incubator-e4-dev] Client/Server Split

[Gunnar Wagenknecht <gunnar@xxxxxxxxxxxxxxx>]

Hi,

One item that immediately tracked my attention is the Client/Server 
split. Is it what I think it is? ;)

A while back I started hacking together some code to "model" UIs. 
However, the model is not as low level as SWT. Instead it's on a higher 
level. The reason is that I wanted to reduce complexity for the 
developers which eventually would use it to model their UIs.

The UIs I'm talking about are UIs often found on typical "back-office" 
web UIs, i.e. pages with forms to fill out, lists and tables to display 
and some dynamic JavaScript effects to avoid complete page reloads and 
to improve the usability.

I made the experience not every developer in a team is comfortable of 
hacking Java, HTML and JavaScript at the same time to build such UIs. In 
fact, only 3 out of 10 developers are comfortable with. That's why I 
started thinking (and coding) this "model".

The most important part of my idea is simplicity. Thus, the model does 
not including a layout requirement. The rendering container is 
responsible for providing a default layout and customization/styling 
capabilities. But they are strictly separated from the UI defined by a 
developer. The developer also should not need to think about data 
binding between server and client.

The UIs are defined on the server. They are delivered to a client. The 
client can be a browser or an RCP application (on a different or within 
the same JVM). Of course, it could also be a DotNET or some other 
proprietary technology talking over a custom protocol. Those 
implementations are just "plugged" into the server. It's one model for 
all "clouds".

The definition for a developer could be as simple as:
https://cloudfree.net/svn/trunk/doc/example1-container.png
which results into something rendered in a browser like this:
https://cloudfree.net/svn/trunk/doc/example1-gwt-rendered.png

Of course, there are many things not yet implemented (eg. field 
decorations, tables, lists, etc.). The first browser "renderer" is 
implemented using GWT.  There can also be a RAP (or an SWT/DOJO?) 
renderer as well. I did start prototyping and RCP Forms UI based 
rendered but haven't had time yet to complete it.

I wanted to reach out to give you something to explore and to throw in 
yet-another-idea for e4. I don't think that such a higher level model 
should be used to rebuild the entire Eclipse IDE. This is out of scope. 
However, I do think that it's something that a lot RCP folks might be 
interested in.

Of course, there are use cases where more specialization is necessary 
which satisfy the need for a full RCP client. Don't think of this 
"model" as a replacement for RCP but as a possibility to share some 
common grounds between the RCP and Web world.

If you want to have a look the code please don hesitate. It's licensed 
100% under EPL.
https://cloudfree.net/svn/trunk/

-Gunnar

--
Gunnar Wagenknecht
gunnar@xxxxxxxxxxxxxxx
http://wagenknecht.org/