drupal

Not long before my old laptop got to the end of it usable lifespan I started playing with the Zend Framework in my spare time. One of the cool things about ZF is that it wants to use friendly URLs, and a dispatcher to handle all the requests. The downside of this approach, and how ZF is organised, it works best if you use a Virtual Host per project. At first this seemed like a real pain to have to create a virtual host per project. One Saturday afternoon I worked through the apache docs and found a solution - then I found it fantastic. Rather than bore you with more of my views on Zend Framework, I will explain how to have a virtual host model that requires a little work up front and is very low maintenance.

It gets tedious copying and pasting virtual host config files each time you want to start a new project, so instead I let Apache do the work for me.

I added a new virtual host config file called projects to /etc/apache2/sites-available. The file contains

UseCanonicalName Off

LogFormat "%V %h %l %u %t \"%r\" %s %b" vcommon

<Directory /home/dave/Projects>
Options FollowSymLinks
AllowOverride All
</Directory>

NameVirtualHost 127.0.0.2
<VirtualHost 127.0.0.2>
	ServerName projects

	CustomLog /var/log/apache2/access_log.projects vcommon

	VirtualDocumentRoot /home/[username]/Projects/%1/application/www
	AccessFileName     .htaccess
</VirtualHost>

The important bit is the VirtualDocumentRoot directive which tells Apache to map a hostname to a path. I use an IP address from the 127.0.0.0/8 range for the virtual host, so they aren't accessible to the outside world and I don't have to worry about it changing every time I check locations.

All of my projects live under ~/Projects and each one gets a directory structure that looks something like this.

[projectname]
  |
  +- notes - coding notes, like grep output when refactoring etc
  |
  +- resources - any reference material or code snippets
  |
  +- application - the code for the project
     |
     +- www - document root for vhost

There are usually other paths here too, but they vary from project to project.

To make this work there are few more steps. First enable the new virtual host

$ sudo a2ensite projects

Don't reload apache yet.

Next you need to add the apache module

$ sudo a2enmod vhost_alias

Time to edit your /etc/hosts file so you can find the virtual hosts. Add a line similar to this

127.0.0.2 projects phpgw-trunk.project [...] phpgw-stable.project

Now you can restart apache

$ sudo /etc/init.d/apache2 reload

This is handy for developing client sites - especially using drupal.

Now my /var/www/index.html is just an empty file.

I am getting a bit bored with adding entries to /etc/hosts all the time. If I get around to adding dnsmasq with wildcard hosts to the mix, I will post a follow up.

This setup is based on my current dev environment (Ubuntu Hardy), but it also works on older versions of Ubuntu. The steps should be similar for Debian and derivatives. For other distros, it should work, just how to make it work may be a little different. Feel free to post tips for others in the comments.

I have been thinking about how to deal with releases of phpGroupWare. For me it is a technical, procedural and political question.

Over the last few months I have been playing with drupal a fair bit. I love drupal. It is simple to install, skin and hack. The community is great. The website is massive and has almost anything you want about durpal. They dog food their stuff. I have quite a few clients using drupal for their sites, they love it. There are many cool things on technical level within druapl too - but that would take this post off on a long tangent.

Drupal was allocated 20 summer of code places by google. phpGroupWare received 1 spot, indirectly. To me this is a sign of the popularity of the project and the level of activity within the community.

I hear you thinking, but hang on, drupal is a CMS, phpGW is a groupware suite, compare apples with apples. Well, you see I am not trying to compare apples with apples. I am looking for good ideas about how to build a quality release.

I think drupal have it sus'd. The have the core which is released when it is ready. They also have a stack of modules which are released when the developers feel like they are ready. This provides a lot more flexibility to all developers. Developers can prepare versions for multiple versions of drupal, but also release stuff when they think it is ready for release, not wait for the next mega tarball to be prepared.

I think that the drupal release model may work well for phpGroupWare. We could prepare the core (probably API, admin, addressbook, calendar, email, filemanager, notes, preferences, setup, todo - the PIM apps amd sync when it is ready) and release that as phpGroupWare 0.9.18. Then modules developers would be free to package there modules and release them when they were ready. Modules which were tested and stable at the time of the official releases would be listed in the release annoucements. It would mean that getting our apps site working would be important as that would be the entry point for a new eco system. It would also mean that if someone is working on new features for an app, they could release often, while the core would be more static and stable, in order to encourage more app development.

I don't think that we can have a release time table for the core unless we have significantly more (paid?) resources available.

I think that covers the technical and procedural issues, the political hopefully won't be too painful either.

The decision on what is core and what is not will need to be made early. The criteria for assessment should be made publicly available. Developers should be free to ask that their app be considered core. Being a core app is not a vital thing, apps will still be promoted it they are not core.

We could look at 4 levels of apps. Core, as discussed above. Supported, apps which meet the core app standards, but are not considered core for what ever reason. Unsupported, apps which work but do not meet the project standards for some reason (coding standards, bypassing the api, lack of docs, require patches etc). Dead, apps which are no longer maintained and other developers feel should no longer be maintained. The status of the apps would be publicly listed.

Such a model as proposed above would allow us to release a 0.9.18 core with some additional supported apps (such as ged, property, messenger, tts) a lot sooner than trying to get all the 0.9.16 tarball apps ready for release.

The only technical issue to resolve in this plan would be version control as savannah's CVS. isn't really designed for this model of development. I have been discussing switching to SVN with the savannah hackers, they are supportive of the idea.

I do have comments open on my blog, so feel free to leave a comment, otherwise discuss it on the dev list. I will post a summary of the comments there if I feel it is warranted.