Myles Carrick

I use Git and want to track upstream WordPress.

[UPDATE] I was using a utility here called svn2git. That was crazy. Now I'm just using git-svn and it rocks.

Here's what I do to track Wordpress SVN. It couldn't be easier.

git svn clone -s http://core.svn.wordpress.org wordpress_svn

The -s flag tells it you've got a normal-looking SVN repo. I clone it into a git repo (folder) called wordpress_svn

It takes FOREVER... replaying svn history one commit at a time.

When you're done... have a look at the branches

cd wordpress_svn git branch -r

To refresh it (or if it dies halfway through) just

git svn rebase

So I decided this afternoon that with a wife 39.5 weeks' pregnant I'd get into devops fun experimenting with Chef.

My goals are simple:

give me a clear, unambiguous, repeatable way of deploying servers of various types

help me deploy a bunch of PHP stuff I need right now:

WordPress with Network stuff

Drupal Commons

I toyed with the idea of getting someone to do it, but sometimes it's just easier to do it yourself (read: faster and far less complicated).

Why Chef?

The reasons are fairly simple, and all about me:

Chef's all Ruby. I know and love Ruby.

I can use Opscode's hosted platform for the servers. The last thing I want to do is run another server or bunch of servers to control my servers - I'm trying to be quick and lightweight here.

Where am I going to deploy?

I'm not entirely sure where this is going to end up. I have some unused Linodes that have to either start getting used or die. Other possibilities are Ninefold and of course AWS/EC2.

Ryan Singer from 37signals mentioned this vid from Stanford GSB

Pixar founder Ed Catmull addresses the question: "Why do 'successful' companies fail?"

What's more important? Great people or great ideas?

Ed goes through some important things that have made Pixar awesome, like:

Undergo constant review - review it all every day - push through the embarrassment; make it so that 'when you're done, you're done'

Keep it safe to tell the truth

Communication and org structures shouldn't necessarily match - everyone should be able to talk to anyone about anything; managers shouldn't be precious about always being kept in the loop.

Don't let success mask problems

Restructure the postmortem / retrospective - keep it fresh; allow it to be sufficiently in depth.

He also addresses the strange feature re company themes / mottos - they tend to be truisms - everyone knows they're true; all competitors have pretty much the same one.

Sometimes having the pithy statement means we can say it and not change behaviour

There are, Ed surmises, two types of 'crisis':

you don't like what you see & have to make changes. People will have to change. It's hard.

you finish and it's rubbish (game over)

The first sort is a LOT better - it's self-imposed. It's harder. But it might actually work.

Fellow contrarian Mark Smithers writes on the phenomenon of widespread, automated lecture capture in education›:

the large scale implementation of lecture capture is probably one of the costliest and strategically misguided educational technologies that an institution can adopt

At UNE we're currently in the throes of deciding our plan of attack for all things rich media. Naturally this includes lecture capture, but the peculiarities of our environment have thankfully provided the context for a re-think. In particular:

80% of our students study all or mostly online

An ever-increasing number of our teaching units have no face-to-face cohort - i.e. no actual lectures to record

The tools for academics to record great, focused learning interactions are inexpensive, powerful, and much easier to use

We don't have the legacy of an existing lecture capture practice

As Mark has highlighted, a gazillion studies have shown (and our own experience often attests) that an hour-long, lecture-style video is rarely the best way to engage, stretch, develop and enhance learning.

So what are we going to do?

We're currently working at ensuring that our focus is on (a) the tools our students tell us they want to use... and (b) those our great teaching staff are keen to get their hands on.

Thankfully, those two are one and the same:

We're rolling out great synchronous meeting / classroom platform. Based on Adobe Connect and tightly integrated with our Moodle site. It'll give us a great building block for fantastic online discussions - for tutorial-style discussions, supervision, collaboration, and more.

We've put in place the absolute best media storage and management platform. Kaltura forms the foundation for this - we've started with terabytes of storage and are sponsoring the open source development of the Moodle 2.0 plugin. It'll be possible to record or upload video straight into the browser... and Kaltura will automatically take care of the formats, resizing and distribution.

We're investigating the best way to give all staff (and potentially students) the easiest way to capture audio and video from the desktop and in lecture-style situations.

OSX Snow Leopard has a command line zip/unzip utility...

Today I needed to unzip a MASSIVE (10GB) zip file (for Adobe LiveCycle - but that's another story)... and got this response:

unzip LiveCycleES2.5-0.0.5.zip Archive: LiveCycleES2.5-0.0.5.zip skipping: LiveCycleES2.5-0.0.5/LiveCycleES2.5.x86_64-0.0.5.vmdk need PK compat. v4.5 (can do v2.1) creating: LiveCycleES2.5-0.0.5/ inflating: LiveCycleES2.5-0.0.5/LiveCycleES2.5.x86_64-0.0.5.vmx

So what's PK compat. v4.5 ? The standard zip libraries don't let you unzip really big files. The PKZip libraries have evolved with myriad different versions and capabilities - including (in 4.5) the addition of Zip64 - with support for huge (>4.5 GB) files.

To get it to work, we need p7zip the Linux/Unix port of the 7-zip libraries.

I use homebrew - and you should too.

brew install p7zip

then use the 7z utility (with a magic 'x' for extract like so:

7z x LiveCycleES2.5-0.0.5.zip

I use Compass with Blueprint for all projects these days.

When deploying on Heroku, the read-only filesystem has trouble with the conditional around the IE stylesheet - leading to something like this;

Errno::EROFS (Read-only file system - /disk1/home/slugs/.../mnt/public/stylesheets/compiled/ie.css - Heroku has a read-only filesystem. See http://docs.heroku.com/constraints#read-only-filesystem)

I get around it by asking Compass / SASS to not bother trying to regenerate my sass/scss files in production:

in config/environments/production.rb:

Among the various things we're doing for UNE's move to Moodle, one thing we thought would be cool would be to remove the ancient option Assignment type that allows only one student file per submission (and no teacher "response" files).

This gives us a stack of simple benefits:

we only have to hook up online marking to one type

we can hook our funky new print and scanning system up to one type

teachers never get stuck having allowed one file and then wanting to allow more

teachers don't have to remember that the more limited type doesn't allow them to provide responses

Making the changes is really simple:

First, add a config setting (in config.php):

$CFG->assignment_hide_uploadsingle = true;

Then tweak mod/assignment/lib.php as per the commit at https://github.com/une/moodle/commit/cf89f14ffc36cae22175874f6f737f4e773cfd74

The change is basically:

- $standardassignments = array('upload','online','uploadsingle','offline'); + $standardassignments = array('upload','online','offline');

Moving to Moodle in 2011 brings with it a number of challenges for how we manage assessment. It also introduces a stack of exciting opportunities.