Having argued that we need to define Open APIs better I’ll share my experiences as a contributor to Open Knowledge and the challenges that this poses. The first message:
Openness takes effort
Creating copyrighted works costs no effort. Every keystroke, every digital photo is protected by copyright. Whereas to make something Open you have to specifically add permission to use. This is really tedious. It’s particularly difficult when the tools used have no natural support for adding permissions. So
Most of the time I fail to add explicit permissions
That’s a fact. I am not proud of it. But, to give an example I have just discovered that almost all my material in Cambridge DSpace does not provide Open permissions. That was never my intention. But the tools don’t allow me to click a single button and change it. I have to add rights to every single document (I have ca. 90). Meanwhile the automatic system continues to pronounce “Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.” This may be true, but it isn’t the sign of a community trying to make material Open.
I have spent the last 20 minutes trying to find out how to add permissions to my material. It’s impossible. (DSpace is one of the worst systems I have encountered). So, unless you get everything right in DSpace at time of submission you cannot edit the permissions. And, of course, by default the permissions are NO, NO, NO. I should add that some of the material has been uploaded by my colleagues. So next rule:
Unless EVERYONE involved in the system understands and operates Openness it won’t get added.
So all my intended Openness has been wasted. No-one can re-use my material because we cannot find out how to do it.
Now take this blog. I started without a licence. Then I added an NC one. Or rather I asked the blogmeister to add it (I was not in control of the Blog). And that’s the point. Very often you have to ask someone else to actually add the Openness. Then I was convinced of the error of NC and changed to CC-BY.
However in later 2010 the sysadmins changed the way the blog was served. This caused many problems, but among others it destroyed the CC-BY notice and licence. So:
Sometimes the systems will destroy the Openness.
So I would have to put in a ticket to have the CC-BY licence restored. All these little things add up to a lot of hassle. Just to stay where we are.
So to summarise so far. (Remember I WANT everything to be OKD-Open unless indicated).
- Blog. Not formally Open. Needs a licence added to the site and to each post?
- Dspace. Almost nothing formally Open. Unable to change that personally. Would have to spend time with the repositarians.
- Posts to lists. I *think* the OKF lists have a blanket Openness. But I’m not sure.
- Tweets. No. And I wouldn’t know how to make them Open.
- Papers. When we publish in BMC or PLoS they are required to be Open and obviously are. Pubs in IUCr are “Green” because we publish on CIF and this is free.
Now some better news. All our software is in Open Source Repositories.
- Software/Sourceforge. Required to be Open. Licence in (some of) the source code. Probably LICENSE.txt indicating Artistic Licence.
-
Bitbucket. Ditto.
Open Source software Openness is fairly trivial to assert.
Services.
- OPSIN. (http://opsin.ch.cam.ac.uk/ ). This is a free service (intended to be OKD-Open, but not labelled) which converts chemical names to structures. Software is Open Source (Artistic). Data comes from user. Output should be labelled as Open (but isn’t). Would require Daniel to add licence info into website. The primary software (OPSIN) is Open Source. However I expect the webserver is the typical local configuration of Java/python/Freemarker etc. and doesn’t easily transport. So what do we have to do about glueware? If it doesn’t port, is Openness relevant?
Services are naturally non-Open by default.
Glueware is a problem.
Data.
-
Crystaleye. 250,000 crystal structures (http://wwmm.ch.cam.ac.uk/crystaleye ). We worked hard on this. We have added “All data on this site is licensed under PDDL and all Open Data buttons point implicitly to the PDDL licence.” On the top page, “All of the data generated is ‘Open’, and is signified by the appearance of the
icon on all pages.” At the bottom, and the Open Data button on each page. And yet, according to CKAN, it’s still not Open because it cannot be downloaded in bulk. (Actually it can, and Jim Downing wrote a downloader. This was non-trivial effort, and I’ll come on to this later). So this is our best attempt (other than software) at Openness.
Even when the data items are OKD-Open, the site is not necessarily so.
So here we are, trying to be Open, and finding it a big effort, and failing on several counts. I bet that is generally true for others, especially if the didn’t plan it from the start. So
Openness has to be planned from the start as part of the service.
(The trouble is that much of what we have done wasn’t formally planned! Much is successful experimentation over several years).