Middleware Matters

Hi Steve,

Excellent article, as usual. I wonder a bit though about the REST pointer at the end, which seems like somewhat of a non-sequitur ... If you look at what Savas Parastatidis and Jim Webber call MEST, or in fact any MOM-based solution, these would qualify just as much. Or am I missing something?

Stefan

Hi Stefan, thanks for the comment and the compliment. There's a ton of stuff I could have put at the end, as you point out, but I was primarily limited by column space (and column deadline). I wanted to mention the Axis2 stuff because it shows the evolution of that particular toolkit in the WS space to also incorporating something REST-like, so that's why I chose to mention it alone in the limited space I had left at the time. In fact, Doug Lea, who is in charge of all IC columns, suggested while reviewing a draft that I consider doing a part II to this column to talk about all the other things I could talk about.

I think that is not a good decision to try to fire RPC. Both RPC and MOM middlewares are needed because each one has its field of application. The world is both synchronous and asynchronous, not only one of those. There will be always situations where you have to wait for an RPC result before to continue.

I think that the question is not if RPC should exist or not, but if the current WSDL approach of modeling web services is good for modeling RPC web services. I think that not, and in that direction I will be preseting a paper next week on the MWS 2005 workshop (http://elab.njit.edu/edoc.html) entitled "WSODL-An object oriented specification for RPC web services".