San Mehat On Web Services & .Net
A reader writes: "There's an interview with San Mehat in regards to .Net & Webservices. He has some interesting comments about what will work and what won't work, and where things are going." San is well known for his Netwinder work, as well as being a good DJ. And, in the interest of full disclosure, San does work for VA Software, the parent company of OSDN, as is DevChannel.
I think it would benefit Microsoft if they made the framework for .NET open source. The dedication and expertise of the Open Source developer community would greatly enhance the reputation of .NET, leading to wider global deployment.
Is that Visual Studio eats a monstrous one. It's really hard to develop (in VB.NET - ack - not my choice) on an IDE that is buggier than an ant farm. Every time I build it's basically 50/50 whether or not the compiler is going to start throwing spurrious exceptions.
Give me Java, or give me death.
Microsoft pushes it all a little bit too much. It's really not worth it.
This is RiverTonic's sig.
.NET isn't that bad and VS.NET isn't that bad. That being said...I'd rather not use VS.NET. I've never been comfortable with it to be honest. ASP.NET has made my web stuff so much easier it isn't even funny. I used to be doing PHP stuff and then tried ASP 3.0. I never really liked either of them...I'm kind of excited to see where this stuff goes. And as for the post on VS.NET being buggy...it's not.
Do you mean alternatives like CORBA, or like REST? REST to me seems the proper way to go about web services for 99% of web services people are building. Most people are doing simple calls... the only trick that remains (and is evidenced in the interview) is a simple means of creating objects that represent web service calls and results, to make working with the calls more natural in the OO language that most corporations are using right now. I'm hoping a simple mapping layer on top of a pull parse is a good answer - I'm trying out JiBX for that although it's still rather beta.
In theory with a good mapper to and from the XML should alleviate the collection problem they talked about in the article by naturally generating good XML for Maps and Lists, and converting back just as easily.
"There is more worth loving than we have strength to love." - Brian Jay Stanley
But at the same time it's not so easy to pass a serialized type from one system to another without rolling your own solution. Int and String will have no problem but what happens when you try to pass a custom collection type that derives from a hashtable in .net to a j2ee system.
I agree with what he said about it not being spec'd out but I also agree with you in saying it should be left out of the SOAP specs. I think that 3rd party software similiar to Borland's Janeva will come into play when interoperating between two different systems. But a few more complex types in SOAP would be nice.
But, we live in a world where STL is a normal thing. If you're a C++ or Java programmer or any kind of an object-oriented programmer, you must have some semblance of containers
I don't think .net has generics yet.
The questions: Mono is the .net runtime/compiler/interpreter for C# (yet). But what about the proprietary code ? the windows forms etc ? All the .net apps that have a MS-based gui will not be allowed to run in mono. Will they ? And how will mono handle those Win32 calls ? Maybe through wine ?
I've always found that SOAP had all the markings of a specification developed by a committee that want to make sure that everything made it in to the first draft. Personally, I prefer XML RPC : http://www.xmlrpc.com/
Jason.
Indeed, I'm working on a major info integration effort spearheaded by GM, Ford, Chrysler, Toyota and others revolving around credit apps. The core of it is just as you describe in that SOAP is being used simply as an envelope. An extensive schema has been developed for the payload using bits from W3C, the Open Applications Group and Standards for Technology in Automotive Retail.