Slashdot Mirror
Ask Slashdot: One Framework To Rule Them All?
New submitter ittybad writes "I work with a small web-based company, and, for some new web applications, we are looking to possibly change frameworks if it will be a benefit to our developers and our customers. We have experience with PHP's Symfony 1.4, and are not happy with what we are experiencing with Symfony 2.0. We have some Ruby guys who would love us to implement a Ruby on Rails solution, and our backend is Python powered — so maybe Django is the way to go. So, I ask you, Slashdotters, what web framework do you find to be the best and why? Why would you avoid others?"
One tool to rule them all: Assembly.
To offset political mods, replace Flamebait with Insightful.
If you have an existing base of PHP and Ruby developers then Cake sounds like the way to go to meet them both in the middle so everyone can pick it up fairly quickly. Cake is based on many of the same concepts as Ruby on Rails so everyone should be fairly at home. It is still PHP though so it won't force all your dev team to write better code as much as RoR will. The flexibility of a PHP base can be a plus though unless it is put in the wrong hands.
http://cakephp.org/
Personally I am struggling through my first Zend Framework Project at the moment but I am not sure I would recommend it as it has caused me a few too many frustrations. I do worry that this will just knock all the other PHP Frameworks into the long grass though as it is by the same people as PHP. I am starting to see quite a few job offers coming my way now I have added Zend Framework to my CV so it does seem to be very popular for some reason.
I just noticed you also mention having a Python powered backend, this may change my advice above but it does bring about another question: Do you really need so many different technologies? Surely this must drive up your costs considerably as you need developers with a much wider skill set or more of them.
I dont read
Drupal can be used as a framework
It's hard to believe that's how Micronians are made. Why don't we see it right now by having you both kiss one another?
If people in your group already love RoR, it's best to go with their expertise. Technically, there isn't enough difference to make it matter.
Backends are virtually always in a different language than frontends (not that that's a good thing, but it shouldn't worry you too much).
NB: The message above might reflect my opinion right now, but not necessarily tomorrow or next year.
http://xkcd.com/927/
Wt is the best one I have tried. I use the C++ version, although there is also a Java version (JWt).
What makes Wt unique is its approach: widgets. You develop web applications like you were developing desktop applications. Also, the API is Qt-like (but using Boost).
I gave up on Rails after I used Wt.
Want a virtualization console? Take Wt, libvirt and an HTML5 VNC client and you are done.
Need Active Directory authentication? Wt, Samba (or Windows APIs if you are on Windows), done.
Streaming? Wt, ffmpeg libraries, done.
Forgetting about bindings and being able to use the millions of C/C++ libraries out there was a huge relief.
My personal framework of choice is CodeIgniter, though if you have Ruby people on the team then you should definitely check CakePHP.
I like CI because it works, and it isn't as arcane as most of the other frameworks out there, meaning that if I want to write my own library, it is fairly easy to do so and I don't need to spend weeks digging through all the other crap first.
Also, it is fairly well documented, and that's very important.
Assorted stuff I do sometimes: Lemuria.org
Why not stay with Symfony 1.4? It's a mature and well-supported framework. We have been playing with Symfony2 ourselves at my current job, but decided to keep using 1.4 until the formgenerators of Symfony2 are a bit more mature.
Of all the php-frameworks i've worked with, Symfony 1.4 still makes me most productive.
I couldn't disagree more. Cake is loaded with deeply awkward black magic and bad practices. Not to mention the fallacy that the model layer is the orm (hint: in the rest of the world it is not). Cake is second on my list of frameworks to avoid (and most senior developers that I know agree). I would suggest you do the same. .
If a man isn't willing to take some risk for his opinions, either his opinions are no good or he's no good
Agreed. Frameworks are nice, but I'm finding them to be very very very overused. Take a minute and really look at your project. Does it actually need a framework? Does the use of a framework save enough time in development to justify the additional overhead? If so, is that because you (or the people working on it) have been taught frameworks as opposed to learning actual programming (laugh if you want, I've met far to many people who know a bunch of frameworks, but couldn't write the most basic raw code if their lives depended on it), or because it actually streamlines the development process? The majority of the projects I've worked on haven't actually benefited in any way from the use of a framework when they've been properly evaluated. Not saying yours is the same, but make sure you take a good, long, objective look at it before you decide on something. My $.02, take it or leave it.
Don't use Cake. There's limited support for actually getting back true objects with their ORM, which means you can't really deal with an intelligent data object. I did a lot of heavy research on the subject last year for my web company and found that Yii Framework (http://www.yiiframework.com) really fit my sweet spot well.
My legacy code is/was all in PHP (up to 8 years of code), but I wanted the flexibility and advantages of a good object based, MVC system that I could fit over top of my legacy code and upgrade as I had time (without having to do an entire rewrite of the code from scratch).
If you mainly do small one-offs that don't require much ongoing work / maintenance, then either RoR or Django would work fine. But if you already have a sizable code base, the benefits of using a framework in the same language is noticeable. I keep finding new things I can do with Yii after a year that make me faster and faster. Haven't run into any needs that haven't already been planned for in the framework (compared to CFWheels, a Coldfusion Ruby-on-Rails clone I've been switching a client to, that while quite thorough, does have limitations I'm already hitting after a week). And the Yii forum is quite active and seems to have steady readership and input from the main committers. So wrinkles with the framework get resolved on a timely basis. It's been a joy coding in.
Good luck!
Correctly flaked, flint is MUCH sharper than any straight razor you can find. We don't use flint in for example atomic force microscopes as tips (instead of say steel) for nothing, you know.
You may think stone age people had just stones for tools, but their flint blades were sharper than any steel (or even soft iron) blade you can get.
I find that comparing assembly to flint is extremely apt. Especially so when you consider the fact that our circuits use a silicon substrate. You do know what flint is, right?
I second the bid for MVC. I've always found the IIS/sql server/.net pyramid to be powerful, stable, and easy to use. Plus it's easy to integrate with all sorts of newfangled toys like mongodb.
There isn't an easy answer. All frameworks are great at getting you 80% done then make the last 20% nearly impossible.
What you know best is properly the right thing to use as long as it's capable of getting the job done and you can still find new staff who have some knowledge of it.
There's no framework for serious development, but you'll do fine with any of the bunch as long as you stay a small timer.
So, save yourself an effort if you plan getting big and go with C directly -- Not only will you earn a magnitude of performance over your competition, your development time will be less as well. That's because competent C programmers will not dick around with completely useless cargo cult programming, like picking frameworks. Nor will they spend 90% of their time building useless OOP frameworks (with 10% of actual functionality) on top of what ever framework they started with.
Keep it simple, STUPID!
... unless you're in the business of throwing together form-based database apps quickly.
That's really all they do well, and there area lot of form-based database applications in the real world, so that's not a small niche.
But for anything that's a little different, you end up spending a lot of time learning the framework, and then even more time working around its limitations. The better approach is to look at your problem and find a set of libraries that are well suited to the task at hand, well documented, well supported, and modular.
Also, take the time to learn your tools properly and exploit them to the fullest. Learning a framework takes time. So does learning about Apache modules and SQL stored procedures. The difference is that the time invested in the framework isn't generally applicable to other problem domains, while Apache and SQL are everywhere, and are worth learning well.
It is tempting, if the only tool you have is a hammer, to treat everything as if it were a nail. - Abraham Maslow
My experience with python-based frameworks is that they tend to help at the beginning and get in the way when you want to do something that is outside what they do easily. Here's what I have learned:
1. If your developers have access to the file system, then stay away from anything that tries to be a content
management system (I'm looking at you, Zope).
2. Think hard about how user permissions will be handled, because if you screw it up it will make debugging and
security a nightmare.
3. Debugging is harder with web-based development than with desktop development. Make sure your framework
has great debugging tools which (for python development) means:
a. The stack traceback is readily available and
b. The framework doesn't try to catch and handle everything. If it does you will find that your error
messages are raised no where near where the actual problem lies and you will have a terrible time finding them.
4. Maybe skip the framework altogether and instead use individual tools. I use:
- webpy for the dispatcher
- Tryton (with Proteus) for handling the database (This allows me to quickly assemble the "administration"
portion of the application in Tryton instead of building a web front-end)
- genshi for templating
- formencode for validation/user error messages
- pyjamas plus YappyCat for AJAX.
Is it sad that everything I have learned about using frameworks can be boiled down to a
short slashdot post?
this signature has been removed due to a DMCA takedown notice
I take the opposite view. I've seen developers who avoid frameworks because they were Not Invented Here, and then spend a lot of their time re-inventing their own framework.
That's not to say that some frameworks aren't bloated and hideous and not worth the cost (the original Java Enterprise Edition comes to mind).
We've gone through a number of options where I work, from a homegrown framework to Struts to various other trials. Eventually, when I managed to pull us into the RIA world, I make a suggestion that got crooked looks initially but which now, a few years on, is seen as ideal: NO FRAMEWORK AT ALL!
We're primarily a Java shop, but this can apply in any shop since there are similar options available for .Net, PHP, whatever else, but I'll describe our model because its very simple: our apps are nothing but POJOs (Plain Old Java Objects for those not in the Java realm) that we talk to via DWR.
That's it.
No Struts, no JSF, no Spring MVC, not even straight servlets! Nothing but pure, simple Java classes with no real tie to any HTTP-related objects (well, usually... some exceptions here and there are required).
The benefits are many: the code is simple and clean... the classes are so easy to unit test that we actually manage to get our developers to do it (sometimes)... configuration doesn't get in the way (no, it's not as simple as some frameworks because there IS configuration, but its so minimal no one minds)... performance is top-notch since there's no extra work being done by a framework first (granted modern frameworks are very efficient and this difference is probably minimal, but still)... and new developers can be brought up to speed in less than a day and no one is ever confused by the code, that's fore sure! There's also been an implied side-benefit: our apps are written in a very stateless fashion since using state becomes unnatural in this architecture (there IS some usage of state used in some places where it's truly necessary, and that's the exceptions I mentioned earlier to not using HTTP-related objects). Yes, this was one of my goals in pushing this approach in the first place, but it's nice that I didn't have to hand down any edicts or anything because it came naturally out of the architecture anyway.
What you wind up with really is a service-oriented design since you're doing more work client-side and the server-side code is a lot thinner... things like navigation and such, transitions between states, are no longer handled by a server-side framework (there's way you still COULD do it server-side, but it becomes pointless). This definitely takes some getting used to and we had our share of paradigm shift-induced ugliness. But we got through it and we're all the better for it.
But, if this isn't the type of application you're looking to develop, if you want the more "classical" web app model, this probably isn't the way to go (although it still can be valuable to mix a technology like DWR in to your, say, Struts-based application... that can be a good first step in fact). You definitely do have to rely on client-side code more (no, NOT at the cost of security, you can be just as robust in that area as you could ever be if you do things smartly). Pair something like DWR with a top-notch front-end library (ExtJS is our choice) and you have yourself a very powerful architecture that you could even call a framework if you want.
My point is simply that you shouldn't get into the mindset that you HAVE to have some big, do-it-all-for-you framework to be productive, and in fact if you go to the opposite extreme and use no framework at all, if you do it wisely, you can find you are more effective then you'd be even with the best framework backing you up. "None of the above" can in fact be a viable and even possibly utopian answer to the original question :)
If a pion (n-) collides with a proton in the woods & noone is there to hear it, does lamdba decay into the source pa
I've read the comments in this post, and I agree with most of them, especially the guys who argue in favor of avoiding frameworks all together. I get where they're coming from, I really do. In fact, not so long ago, I would have made the same argument. The problem is, to do web development, you really need some sort of "framework" or "library". It doesn't make sense to recreate the wheel for URL parsing, environment param passing (ala CGI), etc...
At the end of the day, you'll need to pick some set of utilities in order to be successful at a web project. It's debatable what constitutes a "framework" vs "utility library", because there's a lot of grey area there.
Of all the ones I've used, ASP.Net, ASP.Net MVC, Sprint, Struts, Cake, Symfony, Django and homegrown, I'm landed pretty solidly on Django.
The reason for this is how it really gets out of your way, and just lets you code. It has all sorts of fancy features if you want them, but you aren't compelled to use them.
It's lightweight (I run several Django + postgres instances on a VM with 500mb of RAM with sub 200ms response times), the different parts are pluggable, you can swap out the ORM, templating engine or admin parts for anything you like, and it embraces the pythony way of doing things.
There isn't a bunch of black magic, it's really very straight forward. The framework code is very readable, and minimal. The core "framework" is really just a set of python modules that give you very handy utilities - URL routing, ORM, Templates, etc...
Don't like sessions? No problem, just don't include that module. Don't like the ORM? No problem, roll your own, or use something else.
Want a full blown framework with automatic admin interface, and all the bells and whistles? Great, it's there for you if you want it.
In general, I've been one to avoid frameworks because I agree with the sentiment of many other posters on here - frameworks do the "easy 80%" quite nicely, but the final 20% ends up being weeks and weeks fighting with the framework.
Django is the only one I've encountered that doesn't have this problem. I've never had to fight with Django at all. My only problems were a lack of solid python skills, but one I picked that up, Django was beautiful and made a lot of sense.
It's the most intelligently designed, practical, useful framework I've ever found, and has done what no other framework I've used has done: actually saved me enormous amounts of time.
Drinking habits can be dangerous. You can choke on the cloth and the nuns will wonder where their clothes are.
Could not disagree more. I've worked with a variety of Web platforms/frameworks; on my current job, there is a bit of Drupal fandom, despite almost no one having any experience (except me) with Drupal - it's just become a popular buzzword here (another story).
So one of my first projects after arriving, management had already had it in their heads that it would be Drupal-based. After digging in to the requirements for a week, working on a prototype/proof-of-concept, I quickly hit some walls and realized I'd be spending as much time patching bugs in existing Drupal modules as writing original code - the data model is complex and Drupal's database abstraction layer is about as ugly as they get.
Annoyed and frustrated, after a few beers with an old friend the night before, I read the first few pages of the Django getting started docs on the way home one night - by the time I got home I felt like I had a strong sense for how the framework was structured, the conventions it followed, etc - the docs were clear, concise, and the framework sounded elegant and straightforward, with a clean design (unlike Drupal, which seems to suffer from no particular design).
I hit the ground running with Django and haven't turned back - since that first night with it, I've not run into any big surprises - everything just works as expected. The code is solid, the design obvious, and I'm really in love with Python (having only written simple scripts with it in the past).
I don't think I've ever found the docs to be wanting, and not sure what you mean by the config being touchy - it practically holds your hand, the integrated debug mode gets you straightened out quickly. It does help to understand what Pythonic code looks like - Django is pretty damn close to a perfect expression of what it means to be Pythonic, so it's advisable to get comfortable with Python itself of course.
The one thing I thought I'd hate with Python - the use of whitespace as structure - I got used to very quickly, with the help of a decent text editor. Otherwise it's been a joy.
FWIW.
I don't get it. A framework has piles of domain-specific code that has been tested and is pretty much guaranteed to work. Sure you can hunt around for some open source implementation of every little thing you might use, but having it built in helps.
Chances are, if you're not using some sort of framework, you're using your own libraries or code which basically take the place of a framework. Unless you're doing no output, having something take care of the remaining browser quirks really helps, and especially handling the new mobile platforms.
What quirks you say? Write to the standards? Of course. And then you get a router whose configuration is javascript-based, and does not work at all in Chrome, but works in FireFox and IE. Sure blame Linksys for being crap, or not testing enough, or whatever.
The point is, frameworks do things for you so you don't have to. I can't think of any project that wouldn't be helped from a framework. Unless you're doing something with processing on the server and display on the client. Then it's pretty obvious you don't need a framework, and aren't the audience for the question.