In the recent years I have built my own, mostly basing it on other libraries and reusing as many things as possible.
In Python there's something called WSGI (Web Server Gateway Interface) which makes it trivial to implement your own framework. Around 2 years ago I wrote a blog post about WSGI and how to quickly build a Python framework http://amix.dk/blog/viewEntry/105 In the recent years WSGI has grown a lot. Check out wsgi.org for more info.
One can learn a lot of things by making and using own stuff, but it can also be very frustrating (because of bugs and lack of resources). But generally, I like to use own my stuff and my own conventions - I know at least who to blame when some stuff does not work or something is "ugly" :)
Django, because it doesn't have so much of that special magic. That's why I like it, it works and you know why and how it works. I have no idea if that makes sense. shrug
I like Python and I like the way Django does controllers (not so much the template engine though). But I cannot use Django or any Python, Ruby or PHP setup in my current project because I'm keeping a lot of data in memory and that doesn't work with multi-process architectures.
So I'm using a very small dispatcher Servlet to pass JSON messages on to POJOS and do all the templating in the browser using jquery.
I'm bored to death (and rather annoyed) by Java so I'm always looking for alternatives that support my architecture.
Groovy has no library. It's basically an alternative syntax for Java. I like the way many of Python's libraries work (even though they are incredibly inconsistent)
I used to be pretty big into Zend. I didn't like ActiveRecord in Ruby, or it always felt too confining, and liked the additional control over my models that Zend gave me.
I've recently become a Django convert (not for the pony, that was new to me just now) because the framework just works. I haven't quite worked out the magic required to stand up a Django/MySQL/Apache/Mod_Python server on Windows yet, but it's working perfectly in my VM, so I haven't pursued it very far yet.
If you do it yourself, does it count as a framework?
Server: python, postgresql, cherrypy, simplejson
Client: Javascript, HTML, CSS, Dojo, AJAX.
All the HTML is loaded up front and then AJAX for all the rest. Dojo insulates me from browser quirks. TDD Python is what I use at work (and love) and is good for complex algorithms (language morphology, in arabic).
I don't use an all-encompassing framework because I need total UTF-8 support and I feel more comfortable with the general tailorability of the lower-level libraries over the likes of Django, Rails.
Its greatest strength are all the "widgets" i guess you'd call them. However they are all not created equal. Some are well polished and deliver acceptable performance levels and other times they are just demos and hacks.
The Dojo examples page always froze my browser. Even after switching computers and operating systems, it would hang for a while. I never did figure out if it was just me or nobody else seemed to notice.
I use Django, but only for template inheritance and models - I pretty much leave the template language alone. 99% of the data from the server gets passed out as json.
I'm a cakePHP user. It has gotten and keeps getting a lot better as of recent, and I really like the community. I looked into codeigniter after using cakephp for a while and it just seemed to have a lot less to offer. I tried symfony too and it seemed alot more complex, unnecessarily so.
I'm currently dabbling in Rails and Django, but am going to put them down until I finish my current project, since there's nothing in either that I can't do in cake, as far as my project's requirements go.
These are also a lot less portable than the PHP frameworks, which I find is a big deal if ever you want to lengthen your runway with client work. It's MUCH harder to get a quick 2-5k on a small webpage when the client hears "change hosting".
I've been working on a PHP framework that is similar in the 'full-stack' approach of Cake but has performance characteristics of CodeIgniter. It's called Recess and can be found over at http://www.RecessFramework.org
CakePHP for me too. I also use Zend Framework. I like both. CakePHP is great for it's simplicity, but if their conventions don't work for you try out Zend Framework.
rails because, well, its rails. building something out is typically quick and easy.
codeigniter for those cases when i need to get my hands dirty on a lower level and do some non-standard things. its a minimalist framework that doesn't add much bloat, overhead or other things in your way.
The first syntax is a tad longer, but its much more useful; you can define any method with the name of an HTTP verb for that path, and you can also define a filter parameter that will be queried before the path is matched. The filter lets you do whatever pre-processing you want, so you could filter out requests from a specific user-agent for example, or all URLs that use mixed case, or anything else you can think of.
That looks really cool; how's that been working out? Is that in production? If Ruby had Gambit Scheme's serializable continuations, that would be pretty interesting times.
We're not yet using it in production, its only a few weeks old, but we will be in some of the things we're working on.
One of the main benefits is letting us run objective-j and cappuccino on the server, letting us share application data models in the client and the server.
I like codeigniter a lot because you have more control over things. Frameworks that get things done extremely fast are nice, but I like to have more control over things. And Kohana was modeled after codeigniter, but in completely PHP5, which is why I like it the most.
Pylons when I use Python - magic sucks, very easy to customize, fast turnaround and not so gargantuan that it takes forever to wrap your head around.
CakePHP for PHP. Definitely the best PHP framework out there, in my opinion; fast, good installed base made out of all-star clients (Mozilla, Yale, others), not restrictive or cumbersome. Honestly, though, I try not to use PHP.
Seriously, just install smarty and do it yourself. Then somebody can always do a basic edit to the templates and you don't have to weight the merits of the 10000X different frameworks against eachother and get back to Making Money!
Used wicket for a while on xp-dev.com, but moved to a home grown one as wicket was a tad bit too bloated.
But I really like how wicket is component based, and have used it successfully for other project (all intranet based ones though - so don't really have many publicly accessible examples)
WebObjects. I don't like Java, or Eclipse, which makes it an odd choice, but it was Rails before Rails, and still has a lot of advantages over it. The ORM works hand-in-hand with the framework to do proper persistence, and while I rage against the tools sometimes, they're still better than being dumped back at a texteditor.
If only Apple would give it some more love, it'd be killer.
I wrote my own CMS/framework which I started about 8-9 years ago, so I use that. In my main business I sell a "pro" version of it as well as customization/hosting/etc, and naturally it's also the framework I use for my startup too. For those that are interested:
It's in PHP 4/5, GPL licensed, decently clean code (in 8 years there's obviously some cruft at this point :), but also has a few nice things built-in like a full CMS on top of it, multilingual support, a couple dozen modules that save some time, ORM library, and the ability to setup A/B tests just by editing a page, which is starting to come in handy these days.
Some challenges of rolling your own are that you have to write the documentation, and provide support/training/whatever else since developers aren't likely to be familiar with it (unless it's open and gains popularity). Some things to consider when weighing your options... :)
It isn't a horrible idea. Using a ready made framework all the time without regard if it's right for the job is a horrible idea.
For some apps which have a certain level of complexity and require some extra flexibility, you end up wasting more time fighting the framework than getting the job done.
It helps if you have a well defined standard in place, for python, that means WSGI. Coupled with good libraries, it's hard to beat.
Right, but that's not "writing a framework", it's "using existing libraries in conjunction with one another". There's no real "magic", directory layout rules, etc.
Maybe you end up writing a couple scripts to tie them together or something, but it's hardly a framework. And in many cases, yes, it is a better choice than a framework. And still a better choice than writing your own framework.
Erm, that's my opposite opinion, I guess. My (attempted) argument is that one should not write their own framework unless doing so as an exercise in learning. In any other case, using existing frameworks or using the "roll your own library mix tape" strategy is far more efficient.
Basically, I'm saying if you're using a language which has mature frameworks and/or mature libraries which could be forged into an ad-hoc framework (templating, serving, ORM, etc.) it is terrible advise to tell people "just write your own".
Only in the case that you've considered all existing (relevant) options and determined none of them meet your use case (and none can be minimally altered to do so) should you consider writing your own. Anything else is a waste of time and effort.
Right, your new rule is, write you own framework/library if necessary, otherwise use what's there. That's much better than your original rule which ruled out any use of frameworks as no frameworks could ever have been created.
But Im' still not happy :) I'm sure a lot of popular frameworks were created without any necessity, simply because they were fun to make. Some turned out to have qualities that nobody (including its creator) would have been able to spell out beforehand.
I just inherited a massive enterprise application; you know run a brick and mortar company on it. It's not a one-off; a number of companies use it. The stack is
Jboss;
Spring;
Hibernate;
Struts;
XML gluing it together;
Javascript: prototype, JQuery and others;
Flex and granite;
Vestiges of every Java framework fad of the last 10 years. Maybe I exaggerate, maybe not.
It is way over-engineered. Several times as much code as a Django, Pylons, or Ruby solution. I dont recommend going this path at all, even if you are an enterprise.
After I learned to write Cocoa apps for the mac, I realized I needed a similar framework for web development, so I wrote phocoa.
PHP as a language is kinda bumming me out since I've learned about new techniques like functional programming, closures, etc from Javascript and playing with Ruby.
If PHP 5.3 doesn't seem like it's on the right path, I may switch to another language...
JSPs, Servlets and POJOs. Im a complete beginner so im trying to learn the basics first. Im very aware of some Java shortcomings but I like the fact that I can get responses from thousands of books, tutorials, and the web. A bunch of Python frameworks looks very nice though, and the languaje also, but Im required to use Java for now : (
I use Rails for my bigger project, but I kinda like Google's webapp (the default on AppEngine). It is enough, and I assume it works the best of any framework on AppEngine.
One of my projects is a ___location-based role-playing game for mobile phones that we will be launching soon. The entire game engine is hosted on AppEngine and uses the webapp framework.
I use cherrypy + SQL objects + Cheetah, cause that way i choose what components i want to handle all the general components of a webb app(http reqest/respose, ORM and templating) Also i know exactly how things happen so that makes me feel comfortable ... ie as you guys call it less magic :)
More like "Fartisan"! Just kidding, I've been meaning to give Artisan a try. Where are the docs? Give me at least a nice tutorial on setting up a stupid blogging CMS.
Ruby on Rails is my primary framework. I've played a lot with Merb, but I'm happy that it will be merged with rails. In the future I want to explore web frameworks written in lisps (cl, scheme or closure)
I decided to give Helma a go because I like the idea of being able to re-use server side code within the browser and also because it should make it easier for others to extend the web application with plugins without having to learn a new scripting language.
I've been using Helma NG, rather than the stable Helma 1, so my comments are specific to that.
The dynamic nature of the language makes a huge difference. Both Helma and Grails (which you mentioned) feel like a breath of fresh air after working in Java with frameworks such as JSF. It's much faster to get things done, but I do miss the debugger and IDE integration.
Helma NG takes a new approach to managing scopes and I've found it easy to organize code into modules like in Python.
Helma NG is still work in progress, so parts of it keep changing, which I've found to be a bit of a problem. However, since most of it is in JavaScript and the framework is very lightweight, it is easy to wrap around the parts of a module you want to remain fixed.
I prefer Grails' GSP tag syntax to Helma macros, so might try to add support for that at some stage.
I'm using CouchDB for persistence and have been writing my own abstraction layer on top of that, so can't really comment on the SQL/Hibernate module.
Performance seems good, but I haven't run any benchmarks.
The community is relatively small, but very responsive. The documentation is there, but there are few examples. However, most of the code is self-explanatory.
So, to summarize: Helma NG is perfect for what I'm doing, but isn't yet stable enough to be used in place of Grails for a more traditional webapp. Helma 1 might be more appropriate for that.
Like your experience with Helma, Grails has also been a breath of fresh air. My biggest gripe with building web apps in Java is dealing with all the dependencies, setting up a project structure and making all the frameworks play nicely together...with grails none of those headaches exist...all the stuff you'd usually use is integrated nicely into a project and with almost zero configuration. The community is very responsive and new plugins are released on a weekly basis. I definitely recommend any Java developers out there take a look.
Coding in binary on what? Unless you mined your own raw materials to build your machine, someone else has "done your work" for you. Not to mention where you'd get electricity...
In Python there's something called WSGI (Web Server Gateway Interface) which makes it trivial to implement your own framework. Around 2 years ago I wrote a blog post about WSGI and how to quickly build a Python framework http://amix.dk/blog/viewEntry/105 In the recent years WSGI has grown a lot. Check out wsgi.org for more info.
One can learn a lot of things by making and using own stuff, but it can also be very frustrating (because of bugs and lack of resources). But generally, I like to use own my stuff and my own conventions - I know at least who to blame when some stuff does not work or something is "ugly" :)