Slashdot Mirror
ColdFusion Programming Methodologies?
lars-o-matic asks: "I work at a small (dozen people) company doing quite well building small-to-medium sized sites on the ColdFusion platform and the Fusebox architecture (which also has PHP and JSP versions). With our growth, increasing demand for Flash apps, new features of CFMX, and wanting to take on larger projects, we are researching methodologies. We like Fusebox3 for CF but worry it does not leverage the new object-like CF Components, web services, Flash remoting etc. and wonder if some kind of model-view-controller approach would help separate presentation from business logic. And there's structured documentation, re-usability, maintenance and yes, performance to consider. We're happy with the platform, which suits our project scale. We're not (yet) building a Google or an Amazon.com. It's methodology we need. How have the Slashdot CF users out there scaled from 2 to several coders and from little sites to larger ones?"
For CF, FB is about it. Not that it's bad, but you don't really have anything else to point to re: a structured, methodical approach to web development. Mind you, what's there is pretty good, if still a bit sketchy in some areas.
:)
It doesn't leverage whatever OO might be available in CF. It can't, because FB has a history, and the latest CF is, well, the latest. FB will eventually catch up - I've heard the core fusebox team is working on this issue. Timeframe to 'recommended' specs? Dunno. If you're sticking with CF, stick with FB, or come up with something else which suits you better. If you want to migrate to PHP, consider having us come out and give your developers some custom courses in PHP, suited to the topics you need to brush up on. (subtle plug, but what the hey!)
creation science book
personally, i can't stand fusebox. it's an artificial construct attempting to impose order on what is, essentially, a scripting language. a powerful one (i've been building fairly large scale applications in it for ~5 years), but a scripting language nonetheless. it's just not MEANT to have that kind of structure/organization.
or at least, that's what it was LTE CF5. (as a disclaimer, i have NOT worked with MX.)
with the advent of CFMX, it may get better, but most likely it won't be more than a set of rules for including files to simulated separating business logic from presentation, and "virtual code reuse".
as a corollary, i end up developing a set of my own "code management" rules. develop them inhouse, publish the document, and give a copy to your developers when they're new hires. you can customize it to the way your own shop works, and not be constrained by the artificial rules of another development shop. and hey, you can publish that document and call it "FooBox" (or whatever) and pick up some cash.
Quidquid latine dictum sit, altum sonatur.
Well, judging from the # of posts here, either this ain't the place for CF posts or the market isn't too hot. Shame, too. CF4 did exactly just about everything that a developer of small to medium web-apps. needed or wanted to do. CF4 worked very well with Apache. As the market for small to medium web apps went towards the gutter, so did CF, I guess. Turns out the web wasn't as popular among the basic small- to medium-sized businesses in the business world was (and amazingly still is) thought.
CFMX, as far as I can tell, is the Allaire boys (after selling out to Macromedia) and CFMX trying to be the all-purpose IDE for all dev. environments, instead of just doing its own thing, which got it where it was. Those that try to please everybody generally just please no one. Oh, well. It was really fun while it lasted.
(whatever)
I work at an even smaller shop than Lars, and we develop 90% of our web sites using ColdFusion.
For the smaller clients, they don't even ask what programming language we're going to use. We host most of the sites ourselves, but when a client has their own host, we are finding more and more ISPs waking up to CF and providing it.
A lot of our medium-sized clients are getting in to hosting their sites on their own boxes, and they are definitely interested in what software will be running the site. Once the benefits of ColdFusion are explained to them, they're happy to use it. The fact that the server software is so inexpensive doesn't hurt either. We usually also sell them on the fact that the development will cost less, since developer productivity is excellent with CF.
For the larger clients, I have to do a lot of talking. They sometimes run other sites/applications on the same web server, so they are very careful about what to install. That's one reason I'm very happy CFMX will now install on top of Java Application Servers like Websphere/etc. Larger clients also want to know how this will fit in with the scope of their larger development strategy. Is it a good choice for other applications? (usually) Does it run on our platform of choice (usually yes since it runs on windows, Solaris, Linux, etc.) Is there a large pool of CF development talent to draw on? (yes) Is high-quality tech support and training available. (yes)
On the other hand, although we can convince people to use it, nobody comes and asks for a site to be developed in CF. It's just not a buzzword right now. Everyone is talking about Java and JSP. We are moving towards JSP ourselves, but the environment needs to become more robust before we can make the switch. Coldfusion MX will help with this a lot, since it supports JSP as well.
First, Fusebox isn't an architecture. Second, MVC isn't a methodology.
Any good methodology wouldn't be specific to a programming language. A good architecture for Web applications would also not be specific to a programming language. MVC is a design pattern that can be applied to be different architectures and programming languages. About the only specific thing you really need for CF is an application framework. Fusebox is an application framework; it just isn't very good.
Gah. The web is just the WRONG paradigm (can I say that?) to use for complex interactive applications. When you purchase a car, does it come with a "page-based" interface metaphor? Are there "steering-wheel-emulation" frameworks for emulating a steering wheel on a piece of paper? Yet that is what all these MVC frameworks which attempt to emulate an interactive stateful application on the client side propose to do.
It's just wrong wrong wrong wrong wrong. The web is great for displaying PAGES, not applications.
For the complicated applications people are trying to shove on the web, we need a new solution. Something in between a standalone fat application, and completely server-rendered pages (web). Something like cURL, or XULUX, or (choke) XUL + scripting glue.
When you try to add complicated statefullness and interactivity to the page-based server-based metaphor, complexity scales exponentially...it's just crazy. Your app just becomes a Big Hairy finite state machine DSP.
It's 10 PM. Do you know if you're un-American?
i can't stand fusebox. it's an artificial construct attempting to impose order on what is, essentially, a scripting language. a powerful one (i've been building fairly large scale applications in it for ~5 years), but a scripting language nonetheless. it's just not MEANT to have that kind of structure/organization.
d ef ine
There is nothing wrong with scripting languages and scaling. However, it does depend on your design style. I try to use the database to manage over-all structure, and not so much programming code. More on this at:
http://www.geocities.com/tablizer/misclang.htm#
The most annoying thing I found about ColdFusion was it's lack of first-class functions/subroutines and funky variable scoping rules. You can get subroutine-like structures using custom tags or the scripting syntax, but they are just not "full citizens". There are certain things you cannot do with or in them. Thus, one tends to end up with long "main" scripts. I want full-blown subroutines.
On the plus side, it has something that PHP and ASP do not have: named parameters.
I looked at fusebox a bit, but found it not very adaptable. It seemed to force pages into one of a predefined set of categories and I needed a finer control for the more complex pages which did not neatly fit into a category or spanned multiple.
Personally, I would totally overhaul the way many biz-centric web forms are typically handled in web scripting languages. There needs to be a "view buffer" IMO on the server side, and one talks to that view buffer instead of to HTML directly. The view buffer is then echoed at the end of the script task to the client (after being converted to HTML), but does not disappear. It would make development more GUI-like. Complex form validation and lookup fields would be much easier because you don't have to keep redrawing the HTML from scratch each time with subtle changes.
Microsoft's dot-net stuff comes a bit closer, but they admittedly convoluted their approach for speed purposes. This is a mistake for most biz apps. The best techniques and optimization profile for building eBay and building an intranet are very different. MS sold out to the benchmark wars IMO. More about this at:
http://www.geocities.com/tablizer/webstif.htm
Table-ized A.I.