 Thanks Michael and Kia ora koutou. I am Matthew Plummer and until recently I was based in the art history programme at Victoria University and a couple of months ago I moved to the IT programme, the IT services I should say. I think it's a sign of the time that you can actually move from art history to IT and my talk today will really bridge both those fields to some extent looking particularly at an image management system that we've attempted to develop over the last few years. I've added a question mark to my presentation partly in response to ideas that have been circulating around at this conference partly wanting to acknowledge the fact that sometimes it's better to sort of ask a question than assert an answer and that there's an element of subjectivity as to whether the project that I'm going to be talking about is a success or a failure. In fact I think whether this is a cautionary tale or a really enthusiasm and encouraging project is a litmus test for your willingness to engage in these kinds of things. Also just to sort of provide a little caveat that I came on board on this project somewhat after its inception so I'll be using the term we and I to talk about this project sometimes even though I wasn't involved at the stage just for convenience sake. Okay so Taylor maid what's the advantage of tailoring? Well to be reminded of the importance of good tailoring I think we need to look no further than Madman's protagonist Don Draper. Consistently decked out in an impeccably tailored suit, Adman Draper's product pictures are frequently compelling in part because Draper himself is a premium product and invented identity and adopted persona. Draper you see is actually Dick Whitman and the tailored suit acts as the equivalent of a superhero's outfit helping transform a down and out loser into an advertising superstar. From a certain point of view Don Draper might be the most compelling case for the argument that it's the clothes that make the man and not the other way round. Now regardless of whether you agree with me or not on this point I think it's fair to say that the importance of a good fit be it a suit or a software system is difficult to overstate. Now the questions I hope to tackle in this talk are three fold. Firstly to what extent might open sourced options allow for a tailored fit for institution software solutions? Secondly can these options trump off the rack one size fits all offerings and what are the pros and cons of each approach? And to address these questions I'll be using this case study that I've already mentioned earlier one that has been in development at VUW since 2011. When I pitched the idea for this talk earlier in the year my hope if not my expectation was that by the time this conference arrived I'd have a fully operational system to demonstrate but like grand designs good tailoring it seems takes time and sometimes no short measure of drama. So it's probably not too much of a spoiler to reveal that the system while inching closer every day to a demonstration ready state is in fact still in development. To put it more bluntly three years and counting after its inception it might easily be considered a failure. So I was greatly buoyed by Andy Neil's opening remarks yesterday. They seem to point away to a way in which failure might be embraced especially if it was the result of being bold as I believe is the case here. And I was further encouraged by a number of speakers who touched on the benefits of being able to aggregate content across a range of databases something which the system seeks to accommodate. So in addition to embracing failure and being bold another mantra that I've picked up from this conference is keep calm and aggregate content. I made a patch with myself that I'll never jump on this bandwagon but here I am breaking it just for you today. So let's without further ado have a bit of a look at the system and as I'm sure many of you know stories such as developing a new kind of digital system often begin with a technology trigger and this system was no exception. So for a number of years in the Art History Department the resources that we were used were catalogued using Vernon's content management system. But by the time Sarah Kaila was appointed as a visual resources administrator in 2011 this very antiquated late 90s version of Vernon to be separated from later versions was well truly kind of end of life. And what she realised was that because the cataloging system that we're using was kind of antiquated you couldn't copy and paste into certain fields the metadata system was out of date that there was a culture that had developed of ad hoc cataloging and often inconsistent and infrequent cataloging if it was done at all. It also had become a security risk because the server was running on windows 2000 and security patches are no longer available. So there was clearly a need to update and another key piece of functionality that was desired was the ability to publish content to the web from this content management system, something that wasn't available. A result was that we had all these silo collections a whole lot of academics, artistry staff had collections on their personal drives that were unable to be searched by others. There were limited funds to purchase a new system to develop off the shelf option and where there was funding was really in staff overhead so there was some operating expenditure but not enough money to say by the latest version of Vernon's. And there was also additional complexity because there was the perceived need for the system to cater not only to the needs of artistry teaching staff but also the Adam Art Gallery who are responsible for cataloging and maintaining the university art collection. So what we're really hoping to get was a system which could tick both those boxes, be a great teaching tool, learning and research tool and also a museum cataloging tool which had extra requirements like insurance fields and granular access, these kinds of things. Seridan analysis of existing options within the university at the time and while we had a fiction database it was quite quickly realised that that wasn't going to cut the mustard in terms of the functionality that were required. So what was the solution? Seridan architected this rather bold vision to create a bespoke system utilising two open source software packages and develop an interface between them. If only it was as easy as plugging two codes together. And so a little bit more about these two systems that I've mentioned. One was Ended 3 which is developed by James Madison University in the States and it's a digital image database which has quite a wide range of functionality and it's been used for a number of different purposes. And one of the key things that we are sort of attracted to was the fact that it's a content aggregator so that you could pull in different databases and configure them in different ways and it's all open source. Similarly Collective Access is an open source cataloging tool that's quite widely used in the museum sector and has a wide range of customisation we could implement whatever control vocabaries we wanted. And so the idea was to use these two systems, there's no upfront costs, they're free, they're open and let's develop some connectivity between them. So we started to look at some different databases around the place that we might utilise, some much more easy than others and particularly Digital New Zealand Art Store APIs were very user friendly for the developers. There's also the possibility to port in other systems and the functionality of Ended as a sort of dedicated teaching resource was really appealing. Collective Access provided a method in which we could catalog a range of different systems, a range of different information and could also publish to the web and then we just had to find a way to link all these things together, particularly focusing on this connectivity between Ended and Collective Access. So how did we go about this? During the mind that we had no money or little money. So the funding that we got was from a series of learning and teaching development grants, so rather than pulling money from our limited school resources, we were able to tap into the wider community pool of funds for innovative projects such as this and we employed engineering and computer science students. Initially there was a course called Software Engineering 302 that the students worked on the system in class and then we employed some over the summer of 2011-2012 to see what they could do in this space. What they did was that they arrived at a proof of concept point developing the system in their own environment in the engineering and computer science labs and they also did some customisation to improve the functionality, particularly replacing the federated searching that Ended previously used with United Search and Capabilities. And so I guess when I'm talking about tailor made, this is particularly the areas I'm talking about, A, bringing together two different open source software packages to talk to one another and also trying to improve and increase the functionality of those systems, utilising the skills of our captive developing crew. I should also point out that it's maybe obvious but as a university our IT programme isn't really a development programme or of support so we are able to use engineering and computer science units who had those skills. So what were the customisations? I thought this beautiful Litchfield source from Digital New Zealand might be a nice visual marker. I thought the best way to sort of pitch particularly that customisation of the searching was best sort of put in the words of Rodrigo de Silva who's the key student developing the software himself and he writes, the idea behind utilising United Search as opposed to federated search was to increase code reuse and decrease the amount of work necessary to update or add a new search engines to the server. When implementing United Search we extracted the items which were repeated in each engine and placed them in a location which allows more centralised control. This makes full use of the fact that many search engines essentially contain the same methods and also reduces the amount of time spent searching for certain methods when upgrading and so this enabled specific searches possible bridging the difference between available search engines and terms and as he says it's a simple system but it's effective allowing the maximum depth and breadth of search for each engine. Sounds great right? I mean if you think about some of the ideas that Mia was talking about in the closing keynote address yesterday this idea of trying to aggregate content in the most efficient manner trying to get to that highest common denominator this is some of the customisation that seemed to facilitate this and of course being an open source we can also contribute these improvements customisations back to a community the branches are up on GitHub and have attracted some interest from other people in this area. However it's means being plain sailing and teeding issues I would say doesn't really begin to cover it in some ways. So while we've got proof of concept over that initial summer project getting it installed on an ITS production environment proved to be somewhat more problematic and unfortunately the interface that was developed between Collective Access and MDID has been shall we say temperamental at times. A basic overview of the kind of map of the system this is the server which is hosting the two software sources to get this set up took over a year and this was mainly due to various issues that were not entirely privy to. It took an incredible amount of time to get the server up that we could then try and port in the code that the students had developed and during this time we also had three different software resources administrators overseeing the project. It's a half full-time position I was the third of these and so that's presented issues for continuity there have been sort of customisations or emphasis placed on the project over that time that reflect the differing emphasis and biases and sort of priorities of each of those people. To start this project again today Victoria I think would be much more able to accommodate this kind of system and what we're looking to do in ITS now is to have this kind of three tiered model starting with a chaotic sample where I guess somewhat similar to the ECS students work in their labs creating an environment where there's pretty much no security where it's quarantine from main systems there's a chance to experiment, play around, see what you can come up with and then do security testing test different options for integration and try and get this healthy hot house environment with the idea being that a disciplined engine room and enterprise level solution will result and in order to sort of help facilitate this we've also created a new position in ITS and e-research specialist who has some coding experience and is able to sort of help implement this model. However this model wasn't there when we've done it originally so we've been a bit of a victim of being ahead of the curve Recently we decided for a number of different reasons that let's try this out in the cloud let's see how quickly it would be to implement this system in the cloud and sort of use that as an innovation incubator creative chaotic sample type space it took us less than a week to get this running up on Amazon Web Services versus the year that it took to get a virtual machine in the ITS production environment. Obviously we had the benefit of already doing a lot of work to get it into that environment but it's quite amazing how much more efficient and quick it was to do this. Being in the cloud has enabled a lot of flexibility and freedom in that the coders have complete admin rights over the server they can control what sort of versions of programming they're using, what types of operating system they're using and so on and so forth. Those things are a bit more locked down in an ITS production environment It's accessible from anywhere so with the ITS server that was only accessible on VUW campuses but in the cloud anywhere you've got a web browser you can use it and it's useful for troubleshooting because we can compare the kind of issues we're having and the bugs we're having in a production environment versus the cloud and if the same issues are common to both we know we've got an issue with code rather than some other sort of factor. We're also using Amazon Web Services spot pricing which is not particularly reliable but for what we do it's perfectly fine and it's incredibly cheap but however we're still going to face those same issues of integrating the system further down the track. We've also started to think about how might we as a university sort of bring together a number of these different systems that have been proposed or are being developed and Michael gave an excellent talk in the previous session about an object management system that he's proposed as I've mentioned earlier we already have a Piction database which has got many marketing images, architecture and design is developing a different system so we've got this perhaps inevitable application creep happening and we've started to think as an institution how can we limit this or at the very least how can we get all these systems talking to one another and this is one model that my boss kind of spun up where we would basically have a common storage, we would have as much common metadata across those different systems as possible and all these applications would sit on top of that and then would use some kind of search aggregator on top of that to have one entry point and whether that is sort of a system run by the library proposed by Michael or wherever, the goal is to have increased searchability increased functionality and increased access for all the different digital assets that we have as a university and they are many. So where are we fitting in the hype cycle, the much promoted Gartner hype cycle that you'll hear about already and again no doubt, statistically I'm saying we're about here. We've certainly passed the peak of inflated expectations, hopefully veering crawling inching out of the trough of disillusionment towards that slope of enlightenment. The plateau of productivity I was hoping would already be sort of a little closer but as I said, you know, a litmus test, is this a cautionary tale or is this an exciting sort of example of what can be done, I'll leave that up to you to answer. But I think one of the few, one of the really key things I want to sort of pass on are some of the lessons that we've learned during this project and we do things very differently now and so anyone that's interested in going down this path perhaps this will be useful. First of all grand designs take time, you know you get those nice Kevin MacLeod TV shows and they sort of have time lapses in three years sort of going over, I understand that now much more clearly. This is a term that I borrowed from Danelle McKinley who got it from somewhere else I'm sure, open source is free as in kittens, you get them they're great but then you have to feed them. So there's not that same upfront cost but you have that ongoing costs and also questions about how you're going to support these systems. Timing is everything, I think we've been a little ahead of the curve at least in terms of where we were placed as an institution three years ago to sort of accommodate this kind of project. Timing has also been against us in the sense that we've been relying on funding grants which are often announced at the start of the year. We're using students who have this window of opportunity over summer so we've sort of been, I think, hampered a little bit, delayed a little bit by some of those timing arrangements. That kind of covers that. I think this is a really key point and it's a key part of that question mark after Taylor Mae. The more customised you make the system, the more different sort of tweaks you make to it, the more the support becomes problematic. So that's a really key thing to be aware of that if you develop something you have to be willing to support it ongoingly. And so the importance of documentation at every stage of the process is 100% essential and this is something that needs to be stressed to developers and to project admins. So the complexity of those initial requirements is a two-edge sword because it's meant that it's taken a long time to get a system working but I think it also gives that flexibility to mean it's still current, that we can still find ways and uses for the system that means it's not an outdated thing. So maybe it should have failed fast, maybe it's a good thing that it hasn't. And so to sort of finish with, hopefully, we've concluded in the near future and contact details are up there along with some misogynist but quite hilarious slides that I sort of came across during the course of preparing this presentation. Thank you. Have time for a couple of questions. Hi Matt. Don't mention it at all but I know we're looking at it closely with our system that we're having to move to a more modern system. Ours is quite a lot more than yours. 1992 sort of thing. Anyway, but you haven't talked about migration of the current data. That is a huge, huge part for us and I think it can't be underestimated and you didn't really talk about... There's a lot I didn't talk about. Yes, no, good question. So this is another one of the things that was a bit annoying in some ways but also really fortuitous. Collective access where we'll be migrating that data to is just released a new version, 1.4 when I say just at the start of the year which really made mapping that data much easier. So basically we've got all our visual resource collection which is a few thousand cataloged items in an Excel spreadsheet and it's a pretty easy crosswalk and we're just working through that mapping import process at the moment I'll let you know how we go but it looks on paper to be quite a simple process. You can configure that sort of import quite customizably. Given it took three years and you're still just past the trough of disillusion, if you could go back three years would you still tank the heavily customised two pieces of open source things? Together approach rather than maybe an off-the-shelf something configured and compromise a little bit on what you were trying to achieve and beyond the plateau or productivity, yes. I think yeah and this is once again you know this is a system that was architected by someone else. What I would do differently in retrospect would be just to install and implement the cataloging tool and then because one of the sort of upshots is that in that period of development there's continued to be a lack of cataloging so in retrospect it would have been better to install that open source cataloging tool, get that up and running, continue to use it as we had been with Vernon's and then sort of see if we could graft in the sort of teaching presentation side of things. I guess the way that we went made sense at the time which was to see achieve a proof of concept and see how it could be implemented and there have been a whole lot of reasons which I haven't really gone to detail why though that's been delayed and it's partly just having limited funds and having part-time employees working on it but yeah in retrospect there would definitely be things I would personally do differently. Happy to chat with anyone about this later. Yes, it does. I think the sort of the climate has changed a little bit and I know that in the time of that, since that initial inception, I think there's been a much more heightened awareness about the need for these kinds of systems and there are other systems that we're aware of now trying to do that same thing and maybe that this gets cannibalised. We'll use part of this functionality and that more of it sits with the library say for example rather than within a particular school or programme in the university. It's an open question but I hope that these show that there are options that you have in this open source era to come up with your own system and I think in a slightly more resourced environment it could have been done much quickly and perhaps be a bit more sustainable. Thank you.