05.16.06
Informal Institutional Repository Review – Fez 1.2
Fez is an Institutional Repository front-end solution that has come out of the APSR Project at University of Queensland. Running on a Fedora back-end, it is an Open Source project, and is available from Sourceforge.net.
After starting out with Fez 1.1, we’ve just recently upgraded to the new and improved Fez 1.2. And I must say, it’s nice when a new version comes out that actually fixes a lot of the things that it promised. From the investigations I’ve done so far, it seems that most of the blank-screen-itis has gone – a previously frustrating weakness. I feel that with this version of Fez, we’re finally able to get our teeth into the system and find out how it really works, instead of working around the earlier bugs.
My favourite though, is the import. It works! Now, I know that some people managed to get v1.1 import working, but I just couldn’t. Fez 1.2 however, works a treat. There is some backend setup to be done (e.g tell Fez which directory you’re importing from), but then it’s a matter of putting the material into that directory, then you’re able to use the web frontend to do the importing. As a nice extra, Fez also then relegates this importing process to the background, so that you’re free to go on with others tasks. A status section on your ‘My Fez’ area tells you where the import is up to (although mine seemed to complete within seconds). The newly loaded items then sit in your ‘My Fez’ area, awaiting your input. You can then edit them and submit them to the archive. The amount of metadata that is extracted doesn’t seem as extensive as with some other import solutions, but Fez at least defines what sort of item you’re importing, without you having to tell it. It even knows when it’s importing metadata from a GNU Eprints repository (something I haven’t got working yet, but I’ll persist).
Fez also offers a fine-grained access control, using FezACML. This allows you to set up permissions on items, collections and communities, at all different levels through quite an intuitive interface. It’s not always immediately obvious where and how, but it’s there.
The workflow in Fez is quite nice too – for users of other web-submit forms, all the data entry for a given item is on a single page.
There are some bugs that appear (screens full of text instead of an interface) but this doesn’t seem to affect the actual happenings. There are still a few little bugs (images not displaying, error at submission of item), but these generally don’t affect the functioning too greatly. I’ll put more notes up here as I discover solutions. I can’t give a full listing of what works and what doesn’t but in general, it’s a big improvement from 1.1.
05.15.06
Virtualisation and IRs – Part 2
As I mentioned in my last post, with the setting up of our virtual infrastructure, we have gained the ability to create a new instance of an Institutional Repository in a matter of minutes. Customisations can then be applied and the new system can be ready for use within a few hours. Being able to clone a good base machine means that we don’t have to apply little fiddly customizations time and time again – they are already in the master copy, and once copied the changes are already there, ready to be used.
This leads on to other tracks of thought. Being able to create these machines saves us a lot of time and effort. The amazing thing that I’ve found in the last few months is how few people involved in development have heard about virtualization (yes, until a few months ago, I was one of the un-educated). Outside of my own organization (which is just starting down the track to mass virtualization), I have found only a few other large organizations that have already implemented a fully virtual infrastructure. For them, the thought of going back to traditional methods of building infrastructure is laughable. The amount of time, money and effort they save has more than paid for the initial and ongoing costs of software.
An idea that we’re playing around with quite a bit at the moment is the prospect of being able to ship out pre-built Institutional Repositories. Before we knew virtualization, we thought that a software solution which came in-a-box (meaning in this case that it was just a single script or executable for installation) was the way to go. It made the whole install thing so much simpler than having to install various components and get them to play together nicely. But still, each time we wanted to have a new, empty instance, we were faced with rebuilding almost from scratch. Perhaps we could keep the underlying OS, but we would have to be very careful, to make sure no legacy components were left behind.
Now, with virtualization, that’s no longer a problem. So how about we provide an entire pre-built Institutional Repository to our partners? An entire system – including OS, pre-req software, and the IR software itself, completely patched and ready to rock (if IRs rock?). From our perspective, it’s a simple solution – we spend a little while cloning a new machine from the master copy (at the moment it takes less than 10 minutes). Then we put any desired customizations into the new machine (for example, specific customer requirements). If we then use ESXRanger to back it up (using ESXRanger’s fantastic compression algorithms), we then have a full machine that can be shipped to a customer that’s a fraction of its actual size. If it's not an ESX Virtual Machine, then there are other options available for compression.
Of course, there is the issue of the customer then using the machine. If the customer isn’t wanting to go down the virtualization track in a big way, then there are other options – with VMware having released its successor to GSX Server, Virtual Server, the underlying layer on which to run the new IR is free. So the customer has a Virtual Server, we ship a machine (compressed or not), and voila – one happy customer with a new IR. None of the need for weeks of learning how to install and customize the software. A nice solution, and one which I can see being extended out in many other arenas. VMware is already heading this way, with the support of Virtual Appliances. While a pre-built IR doesn’t come necessarily under that category, it is a complimentary use for the technology.
The use of this type of technology really can extend a long way – exactly how far can fill another post at a later time, but suffice to say that for development such as we’re doing, it’s just about perfect.
