Time to start rebuilding GaiaBB

In a life a long time ago in early 2002, we had to move Australia’s largest Volkswagen car forum from EzyBoard, which was distributing malicious ads and hard to get rid of pop ups to our users, to our own forum software. After a product selection, I chose XMB, which was (and is) better than all the other free forums out there, such as phpBB (didn’t have attachments until v3/0!), and others.

XMB was a good choice as it had so many features. What I didn’t know is that XMB was full of security holes. XMB started life as a two week effort by a then 14 year old, who had limited capabilities in writing secure software. If you looked at the OWASP Top 10 and XMB, XMB had at least one of everything. We had XSS, we had SQL injection, we had access control problems, we had authentication bypass, we had … you name it, we had it. Boards were being pwned all over. So I started to help XMB, and soon became their security manager.

The story of XMB’s rise and fall is long and complicated, with many machinations over its history. There were multiple changes of lead developer, of which I caused at least once, something I’m not proud of, nor realistically never covered anyone in glory. The original 14 year old was pushed out before I got there, and there were various stories floating around, but one of the interesting things is that as a result of that, a company thought they owned a free software project. I pushed to have our software adopt the GPL, which was accepted at the time, and it’s the only reason that XMB and all its forks including GaiaBB lasted to this day.

XMB was forked relentlessly, being the basis of MyBB and OpenBB, as well as UltimaBB and then GaiaBB. The late 2000’s were not friendly to XMB, with the loss of not only the main development system, but also a change of “ownership” caused much rifts. Then there was another failed fork called UltimateXMB, which initially was a mod central for XMB, but then turned into a fully customized version of XMB like UltimaBB, but closer in database schema to XMB itself. The last fork of XMB, XMB 2, was a last ditch effort to stop it dying, but it failed as well as the last “owner” of XMB decided to use DCMA take down requests, which is illegal as I owned the (C) of many of the files in XMB, as did others – particularly John Briggs and Tularis. That last remnant of the one true tree of XMB can be found XMB Forum 2.

In 2007, I forked XMB with John Briggs, creating UltimaBB, which life was good for a while, we had momentum and it worked better than XMB during those years. After the loss of the XMB master SVN tree, XMB 1.9.5 was resurrected as an effectively a reverted version of UltimaBB. Then life changed for me, with moving the US and having a child, so we parted ways, which was sad at the time for me as I knew what would happen without a strong and active lead developer. Eventually UltimaBB withered too. I had to fork UltimaBB, creating GaiaBB as I needed to keep my car forum Aussieveedubbers alive and secure.

GaiaBB got only just enough love to keep it secure, but it hasn’t kept up. It barely functions on PHP 5.6 and modern browsers render it funny. The technical debt from the basis of 14 year old “modular” code originally written by a 14 year old (header.php and functions.php are both monolithic and stupidly long), it’s time to call time on it.

So I need to start over or find something new for my forum. As I need to keep my skills up with the modern kids writing software today, I’ve decided to make the investment in re-writing the software so I can learn about modern Ajax frameworks, and have a go at writing back end code. No small part of this is that I want to learn about the security measures as a developer as a code reviewer and penetration tester, you can’t talk to developers unless you know how applications are built and tested, and all the compromises that go into making applications actually run.

GaiaBB

So let’s talk about GaiaBB. Compared to most PHP forum software, it’s pretty secure. It’s got all the features you would ever need and then a lot more on top of that. But it’s spaghetti nightmare. It needs a total re-write. It’s not responsive design in any shape or form. Mobile users just can’t use it. There are heaps of bugs that need fixing. There’s no test suite. Database compatibility is not its strong point.

Frontend decision – Polymer

After looking around, I’ve decided that the front end shall be Polymer as it has good anti-XSS controls and is rapidly evolving. It does responsive design like no one’s business. And because Polymer hasn’t got the cruft of some of the alternatives it will make me think harder about the UI design of the forum software.

Back in the day, we crammed as many pixels and features into a small space because that was the thing then. Nowadays, it’s more about paring back to the essentials. This is critical for me as I don’t have the time to put back EVERY feature of GaiaBB, but as I know most features are never used, that’s not a big deal.

Backend considerations

Now, I need to choose a back end language to do the re-write. My requirements are:

  • Must be workable on as many VPS providers as possible as many do not provide a way to run non-PHP applications without difficulty
  • Must be fast to develop, so I am not interested in enterprise style languages which requires hundreds of lines of cruft where one line is actually required
  • Must support RESTful API
  • Must support OAuth authentication as although I can write an identity provider, I am more than willing to allow forum owners integrate our identity with Facebook Connect or Google+.
  • Must be a entity framework for data access. The days of writing SQL queries are done. I want database transparency.
  • Must support writing automated unit and integration tests. This is not optional

So far, I’ve looked at various languages and frameworks, including:

  • PHP. OMG the pain. There are literally no good choices here. You’d think because I have a lot of the business logic already in PHP that this would be a no-brainer, but the reality is that I have terrible code that is untestable.
  • Go. Very interesting choice as it’s a system language that explicitly supports threading and all sorts of use cases. However, it does not necessarily follow that writing backend code in Go is the way to go as I’ve not found a lot of examples that implement restful web services. It’s possible as it’s a system language, but I don’t want to be the bunny doing the hard yards.
  • Groovy and Grails. I have clients who use this, so I am interested from learning the ins and outs as it seems pretty self documenting and fast to write. Uses a JVM
  • Spring. Many clients use this, but I do not like how much glue code Java makes you write to do basic things. Patterns implemented in Spring seem to take forever to provide a level of abstraction that is not required in real life. I want something simpler.

Frameworks I will not consider.

The few remaining XMB, UltimaBB, and GaiaBB forums need to be migrated to something modern, and that requires support. I don’t have time for support, so I am going to exclude a few things now.

  • Python / Django. I don’t write Python. Few clients use it and I don’t want to be figuring out or supporting a Python web service layer.
  • Node.js. I know this was hot a few years ago, but seriously, I need security, and writing a backend in something that does not protect against multi-threaded race conditions is not okay.
  • Ruby on Rails. I was thinking about this for a bit, but honestly, I’ve never had to review a Ruby on Rails application, so re-writing my business logic and entities in this will not give me more insight than using Groovy/Grails, which I do have clients.

At the moment, I’m undecided. I might use Groovy/Grails as it’s literally the simplest choice so far, and supports exactly what I want to do.¬† That said, Groovy/Grails is starting to lose corporate backing, so I don’t want to use a language that might end up on the scrapheap of history.

What would you do? I’m interested in your point of view if you’ve done something interesting as a RESTful API.

2 Replies to “Time to start rebuilding GaiaBB”

  1. Interesting.

    I’ve just started building a light, secure PHP framework for a website I want to run. I know it’s secure because I’m borrowing a lot from code I wrote in my dayjob that has passed a penetration test with flying colours. And I know it’s light through sheer experience (and I benchmark some aspects) However, I do need to think about how to craft a test-suite.

  2. Hi Andrew,

    I would +1 for Grails for the following reasons:

    – Underneath it’s all spring + hibernate without the headache of DTO and Factories all over the place. I’m using Grails for the service and data tier of a web app, and it has saved a LOT of time dealing with GORM instead of Java hibernate.

    – Having said that, it makes working with Hibernate less painful, but it’s still Hibernate under the covers. Most of the grails docs/tuts online only cover the simplest use cases. Recommend reading Pete ledbrooke’s series on GORM gotchas: http://spring.io/blog/2010/07/28/gorm-gotchas-part-3/
    Another interesting read is: https://mrpaulwoods.wordpress.com/2011/02/07/implementing-burt-beckwiths-gorm-performance-no-collections/ which takes a lot of the magic out of GORM associations and makes them simpler to work with, faster and easier to debug.

    – A huge timesaver is the JSON and XML slurper capabilities of Groovy. Makes consuming and producing those formats not only easy to do, but clear to read.

    – And then there’s the expressiveness of Groovy, especially the collections methods, a pleasure to work with.

Leave a Reply

Your email address will not be published. Required fields are marked *