<body><script type="text/javascript"> function setAttributeOnload(object, attribute, val) { if(window.addEventListener) { window.addEventListener('load', function(){ object[attribute] = val; }, false); } else { window.attachEvent('onload', function(){ object[attribute] = val; }); } } </script> <div id="navbar-iframe-container"></div> <script type="text/javascript" src="https://apis.google.com/js/plusone.js"></script> <script type="text/javascript"> gapi.load("gapi.iframes:gapi.iframes.style.bubble", function() { if (gapi.iframes && gapi.iframes.getContext) { gapi.iframes.getContext().openChild({ url: 'https://www.blogger.com/navbar.g?targetBlogID\x3d8907963\x26blogName\x3dWS-Comments\x26publishMode\x3dPUBLISH_MODE_BLOGSPOT\x26navbarType\x3dBLUE\x26layoutType\x3dCLASSIC\x26searchRoot\x3dhttp://ws-comments.blogspot.com/search\x26blogLocale\x3den_US\x26v\x3d2\x26homepageUrl\x3dhttp://ws-comments.blogspot.com/\x26vt\x3d-792153501087690591', where: document.getElementById("navbar-iframe-container"), id: "navbar-iframe" }); } }); </script>

WS-Comments

perspectives on open-source and web services

Monday, October 03, 2005

Reynolds's SOA definition

in response to Matt's post, which was a response to Reynolds's post, I would say the following...

cajo's response sounds nice to developers, because it's what they want to hear - we don't need to change the way we do things; SOA is just new buzz-hype on what we already do. but it would not be "honest" to say "that SOA means nothing more than separating business functions into routines, just as they have always done."

for example: one of our developers separated his business functions into routines: he made 11 .cfm pages of ColdFusion code, and then 1 page of ColdFusion code with 12 cfinclude tags. even if the only thing you've read is the senseless marketing trash that vendors are putting out about SOA, you know this cfm approach is not SOA, even though it is separating business functions.

Reynolds includes the key distinction at the end of his definition (a smart place for it):

"Each service provides an interface-based service description to support flexible and dynamically re-configurable processes"

the interface-based service description is mandatory. it's what makes an SOA different. the architecture is not based on identification of objects in the system, as with OOP. rather, objects come about from the descriptions of the services that need to be performed. neither is the architecture based on run-time context as the case with procedural includes of multiple script files.

the service descriptions come first, and are arranged into composite, higher-level services with their own descriptions. this is the key between SOA and other approaches to modularity.

I first encountered this in Erl's first book when he talked about the similarity between OOP and SOA - code to the interface. but with WS-based SOA, the interface is described using a standardized format so that the services implementing the interface can be in any language on any platform.

so the good news is that it is indeed another attempt at creating modularity and re-usability in software. the bad news is that if you work under the assumption that that's ALL it is, you won't really be capturing the benefits of SOA.

4 Comments:

At 10/04/2005 10:40 AM, Blogger Matt Crouch said...

>with WS-based SOA, the interface is described using a standardized format so that the services implementing the interface can be in any language on any platform.

>so the good news is that it is indeed another attempt at creating modularity and re-usability in software. the bad news is that if you work under the assumption that that's ALL it is, you won't really be capturing the benefits of SOA.

I envision it like this: the networked/SOA application is analogous to a well-designed operating system, like UNIX or GNU/Linux. Modularity is encouraged by having processes generally be small, and good at one thing.

Inerop is encouraged by having the vast majority of these processes pipe boring ole ASCII To each other.

When I think about it this way, the benefits of SOA seem kinda obvious, and I wonder why it took so long for this kind of thinking to get hot.

Well, your CF guy -- god bless him -- didn't actually separate into routines. Sticking with my analogy, all he did was write 11 shell scripts containing a line, then one shell script that called all those lines in succession.

Actual routines would be snippets that require an input (or react to an event) and generate an output (or trigger an event)

 
At 10/04/2005 2:00 PM, Blogger luke said...

I think you've mentioned the Lx analogy before and I like it.

but there's still a danger of over-simplifying it with this analogy unless we emphasize that for good SOA, the interface descriptions are the basis of the system design.

so to extend the Lx analogy to be more accurate, I'd say:

modularity thru small processes that are good at one thing. interoperability thru standardized input/output exchanges. and really good man pages describing every process.

then good (Lx) SOA is achieved by designing the architecture of the system based on the interfaces of the utility processes, as described in the man pages - not the internals of each program involved.

ie, if you write a perl program or shell script that merely executes a series of lower-level utility programs, you decide how to structure your program by looking at the man pages. you don't look at the source code of the utility programs, although you could.

in that way, you capture the biggest SOA benefits - re-usability (of the many small processes), flexibility (by re-arranging existsing processes interactions), and modularity (by removing context from the small processes, and the composite larger processes).

 
At 10/04/2005 3:29 PM, Blogger Matt Crouch said...

excellent point. WSDL et al. remind me of phpDoc and programs like that ... generate man pages automatically if your code is up to snuff, properly commented, etc.

I have a confession, though: I never got into this habit. I deserve Bad Things (like always being behind deadlines) as a result.

 
At 10/04/2005 3:56 PM, Blogger luke said...

interestingly, when I asked Andi about how PHP plans to support the automatic creation of WSDL that describes PHP functions, he said it would be implemented using certain comment syntax before the function. then a PHP class would interrogate the code for the comments and create WSDL based on the input/output of a function specified to be a web service.

IMHO, ColdFusion does it just about the coolest way possible. you don't even have to include special comments. just make your CFC and then hit it in a browser with a ?WSDL at the end of the URL and it will interrogate the methods in the CFC and create proper WSDL for it automatically. so all code changes are reflected in the WSDL as soon as the change is made. nice.

I know PHP5 has object introspection features, so I think that'd be the best way to accomplish it. but comment syntax is a good first step.

I also never document well. see you in hell.

 

Post a Comment

<< Home