WebServices - Axis

Summary of Apache/Microsoft Interoperability Meeting
March 12-14, 2001

From Apache: Glen Daniels, Jim Stearns, Doug Davis
From MS: (main contacts) Keith Ballinger, Andrew Layman, Eric Andrae

Most of the 3 days were spent testing Apache SOAP v2.1 and Axis against the 5 different versions of SOAP MS has (one being an IE client). Since Axis doesn't have serialization support yet (just Strings) it was limited in what it could do, but for those simple tests that did NOOPs or tested Strings it did ok as a client and a server (just a few minor tweaks were needed).

Apache SOAP v2.1 did much better. There were a few bugs (and holes) that were discovered but Glen was able to track them down and, I think, fix all of them. He's in the process of trying to see if the fixes can be integrated back into the cvs tree, but is unsure about one of the fix's impact on the MIME support - but we'll see.

MS has 5 different versions of SOAP and they've been doing some of their own interoperability testing internally so I think they were able to flush out most(all) of their differences before we got there. However, we did manage to find a bug (or two?) in their code 8-) but overall they had things pretty well covered.

For better or worse, MS is very WSDL dependent. If the industry is headed down the path of basically requiring WSDL then Apache might need to do so as well.

We had two strategy meetings in which we discussed how to improve interoperability testing/conformance in the future, not just between Apache and MS but everyone. We decided to set-up a consortium in which different SOAP implementations can join and test their version of SOAP against the others in the group. The main purpose of the group would be to focus attention on interoperability issues and not necessarily prove conformance to the SOAP spec. While we will have testcases that we "believe" test some aspects of conformance we can not be the defining authority on who is, or is not, spec compliant. All we can do is pretty much help people say that their SOAP code can, or can not, play nicely with others in the group. That being said, there are some definite MUSTs and MUST NOTs in the spec and we will have some tests that test those so we'll be walking a fine line.

We will also group tests based on sections of the SOAP spec. Not everyone will want to implement all sections of the spec but will still want to test conformance based on what they have implemented.

To help this "consortium" we're going to set-up a web site (Jim has already reserved wsinterop.org and soapinterop.org) where people can post their testcases and test results. We also talked about having pointers to "live" servers that people can hit to test their SOAP implementations. It wasn't decided how Apache will work this.
****TODO*****
We need to get someone to volunteer to set this up/host it.

Along the lines of getting interoperability, we discussed showcasing how nicely we're all playing together. 8-) In particular we discussed having a live demo at a conference (maybe NetWorld Interop in May) where people can hook up their machines in our network of computers and join in "the game". "The game" will consist of a fairly simple maze type of game - each server will own a certain number of rooms and clients will be able to walk from room to room examining, and placing, objects in each one. We'll define a set of base SOAP interfaces that people will need to implement and if they do then they should be able hook-in their server and extend the maze.

Glen is working on the write-up of the overall idea and will distribute it once it's done. In order to help things along we agreed to try to have another F2F around the end of April where we will all get together in a room to hash-out the details of the game/interfaces and to actually code it up. I (Dug) have agreed to see if IBM will host it in Raleigh. Glen is going to see if Allaire (aka Macromedia) will be willing to work on the GUI that the main-server will use to display the status of the game.

PDF