I fail quite often. I have tons of ideas which seem like they might work and loads flop. Occasionally I do something that fails to fail and those are the ones people notice. I usually have fun and learn some skills on things that don’t work out so it’s not wasted time. I thought it might be useful to review some of the things which never really worked out.
I started work on a suite of tools to process N-Triples on the command line. It’s been used by me a couple of times, but nobody else. I really thought unix pipeline tools would be useful, but not yet.
Southampton Developer Meetup
For two years I organised a monthly pub meetup for technology enthusiasts. It limped along for a while but without really reminding people they just didn’t show up, and a couple were only a couple of people at which point I called time on it. It was worth trying, but wasn’t the event people really wanted, which would have had more actual content, and that was more than I was willing to commit to.
For a demo see http://lemur.ecs.soton.ac.uk/~cjg/tableview/ and check out how clean the page source is.
This tool modifies the fontsize until the text fits into a specific height and width. Kinda like what powerpoint does. It’s still a bit buggy on single words for some reason. It’s ideal for making HTML based display screens where you don’t know the resolution or aspect ratio.
This monsterous creation is a perl library and command line tool which talks to some of the more common SOAP functions in sharepoint and lets you get at the data. It’ll dump out datasets as CSV and calendars as iCal. We use it internally but I’ve never heard of anybody else getting any value out of it.
This one is a bit of a heartbreaker. I thought this was a game changer but I’ve never got any interest in it. It is built on top of what Ash & I learned from building and rebuilding data.soton.ac.uk and lets you make a templated website with minimal configuration.
I used DBPedia as the initial demo — I thought it was a good chocie but other people have suggested otherwise.
Ah, well, maybe it’s time hasn’t yet come.
Extra types for RDF. Normally you just use the XSD types for RDF but we found it very useful to use a datatype to indicate if a literal contained plain text or a fragment of HTML markup. Extrapolating from that I created the XTypes vocabulary thinking it might be useful, but it’s never caught on.
I never fully got this one working. It was intended to look at an RDF dataset and tell you what the main types were, what properties they had, and what the objects of those properties were. This is distinct from the vocab, it’s instead what is actually in a given dataset. I thought it would reduce the time-to-grok when working with undocumented RDF datasets.
This was an attempt to automate turning the Southampton University Student’s Union meeting minutes into RDF. It looked at the layout of the document, not just the order of words, and gave surprisingly good results for the effort involved but the interest from the Union waned (as it always does with a one year turn over in leadership).
See a demo of They Work for SUSU.
RDF of the Library Jewish Collection
I thought I could map the metadata from one of the library collections into something semantic, but it was just a list of terms which might be dates, approximate dates, names, organisations or anything else and in the end I gave up. It would have been faster to go through 500 records by hand to annotate if each keyword was a date, person, organisation or something else.
This is a linked data version of the European Common Procurement Vocabulary. When any public sector organisation in the EU buys something big it has to go to tender, and these terms must be used. It contains codes for everything from cigarette papers to warships… but not 3D printers as it’s from 2008. As most university research equipment in the UK will have been procured using these codes I thought it would make a good base vocabulary for categorising things on equipment.data.ac.uk but nobody has done anything with it that I’m aware of.
I really felt that this ontology was a missing bit of the linked data web and would help start to enable real open linked data for large events, which could massively benefit from economies of scale in producing tools. I’ve got good use out of it myself, but it’s not been used by anybody else that I’m aware of.
I expected this to really take off. It keeps weekly stats on a bunch of factors about *.ac.uk homepages and I thought the data would be really valuable to university comms deptartments and web scientists, but nobody seems very interested.
I used open data sources to generate 1.6*5.5km of Southampton as a Minecraft map. I thought people would be really interested and it would inspire people to build on top of it, but I’ve had virtually know intererst even after presenting it at a couple of events.
Well, I feel a bit glum looking at that list, although I enjoyed working on every one of them. I have a few successes under my belt, but they come at the price of a vast number of failed ideas, or at least yet-to-take-off ideas.
What projects have you failed to get off the ground? How do you decide when to stop working on them?