[Pipet Devel] client/server communication layer

Brad Chapman chapmanb at arches.uga.edu
Fri Feb 25 09:02:27 EST 2000


Gary wrote:
> Well now this is an interesting concept.  If I catch your drift, you 
are
> proposing that we use a web server to process the xml, and then
> dyanamically generate the web page and send that to the browser. 
> Considering my distate for the development tools available to access 
the
> DOM from the web client, I could go for that.  It does seem a 
_little_
> odd that we use xml to communicate between the client (gtk, web
> _server_, and nli) if we go by this method.  So...

Right, I think the front should just be a web server front that 
communicates with the middle. I _think_ Jeff also agrees with this. 
This way, you can use any language you want to talk with the middle 
(by sockets or corba), get the xml from the middle, and do whatever 
you want with it. Then you can use anything in the bag o' web master 
tricks to serve up nifty web pages. 

Gary wrote:
> So do you then propose to pass xml through corba, or to just drop xml
> for client/middleware communication?  If we go corba, i really don't 
see
> the point of using xml.

If we use corba, I was still thinking of using the same interface (ie. 
send data, return data) and not a really complicated interface (ie. a 
special interface for connecting loci, a special interface for adding 
loci, etc...). Then we will still need to have some way to 
communicate structured information back and forth, and I think xml is 
good for this, since we can then use the tools built for xml, and not 
have to define our own language and parsers (which is what they do in 
the microscope example I keep talking about).

> Why not?

I don't really know enough about corba to argue convincingly for it 
now, so I think Jeff is right, we should start with the simple, and 
see what happens from there.

Brad







More information about the Pipet-Devel mailing list