Document management sucks! There – I said it. I challenge you to prove me wrong.

I haven’t yet found a document management system (DMS) that’s fit for purpose, and I think I know why.

It’s not about the technology. Documentum might hark from the client server era, and Alfresco trumps that with its SOA, but these are implementation details that matter not to the user.

Paper is two dimensional

It’s about the metaphor. Specifically the dimensionality of the metaphor. Pieces of paper are 2D, and so are document management systems. This makes sense in the physical world. I can only put my piece of paper in one place, which I might call a file or a folder or whatever. Computer file systems copy this metaphor, and document management systems copy it again – they just call a folder a workspace to make it sound more collaborative.

The real world is multidimensional

And we have IT abstractions that can be too. When I write an engagement letter to a client I shouldn’t be forced to ‘file’ it in a ‘folder’ called ‘engagement letter’ or ‘clientX’. I should be able to give it multiple attributes (tags), and virtual ‘folders’ can be assembled from those attributes. Thus if I want to see a library of engagement letters I select that tag, if I want to see all of the stuff relating to clientX I choose their tag.

‘Oh’, I hear you cry, ‘tagging takes effort, and people will forget’. That’s a real problem, but I think it’s solvable. We stopped trying to categorise everything into some massive (2D) taxonomy a long time ago, and decided that search would fill the gap. Search is good when it works, but it can be disorderly – hence the glib one liner ‘why are you searching when you should be finding’. The right way to do this is to make search part of the process at the front end rather than the back end. Clippy returns – ‘it looks like you’re saving an engagement letter to clientX – shall I tag that “engagement letter” and “clientX” for you?’.

The Social Aspect

We’ve seen great usage of tags already in social bookmarking sites like del.icio.us, so why not bolt on that functionality to document management (after all a ‘web page’ and a ‘document’ are essentially variations of the same thing). This raises the question of whose tags – my tag, your tag, the company tag. But who cares – this is what educated suggestions can help with, particularly when search can identify similarities with other documents. As JP wrote earlier in the week, Social Objects are important, sadly I fear that document management systems are anti-social. A DMS may provide a ‘shared space’ for a document as a social object, but it fails to provide rich support for the other activities that should be taking place. Yes, there’s metadata in a DMS, but not typically the sort of open and collaborative metadata that JP is referring to. This is where I can get excited some more about initiatives like Open Bookmarks (which is best described here). Not only does this potentially solve the state synchronisation issues I was concerned about last month, but it provides the perfect platform for social interactions around documents – a means to provide curation.

Security

This is an area that can’t be ignored. As soon as you get into a conversation about accessing documents you’re quickly into a conversation about preventing access to documents. I think this isn’t as hard a problem as people make out. Identity Management practices can be applied equally to people and the documents they interact with. People have attributes (like ‘Director’ and ‘works in HR’ and ‘based in ‘Switzerland’) and these can be synthesised into abstract roles (like ‘Swiss HR Directors’). Documents also have attributes (tags) that can be used to provide not just logical views as discussed above, but logical groupings for the purpose of security policies (for example all documents relating to ClientX can only be seen by the ClientX project team, except NDAs and engagement letters, which are also visible by legal). The policy here provides a mapping between document views and people roles, and should be fairly self explanatory (and easy to audit).

Now can somebody please build this for me?

PS This is yet another reason why we shouldn’t have software patents. Firstly this whole thing is obvious (at least to me), and secondly Xerox have this, but in their usual style appear to have done precious little with it. All that’s achieved here is blocking a startup from implementing (or robbing them blind later for ‘infringement’).