Reference managers


I spend a lot of time reading papers. Keeping up with the sheer volume of articles on subjects that interest me is a nearly impossible task. As a consequence as I’ve ended up trying out most of the reference managers out there, and frankly, not impressed. They’ve all failed in some way or another. Sometimes it’s because the software is unstable, Papers 3 suffered badly from this when it was first released and it’s really off-putting—why invest in a piece of software the developers don’t really care about? Others lack features and the development seems glacial, yet others are just difficult to use. There’s been more than a couple of times where I’ve had to go back to the journal or use Google to find a paper I know I collected! That is an epic fail in a reference manager.

Some thoughts on the ones I’ve ended up using for an extended time:

  • Mendeley—This showed a lot of promise early on and was my main reference manager for a number of years. Getting an iPad changed that. The iPad is such a nice device to read papers on, but it took Mendeley a long long time to get a functional iPad app. I think the 3rd party PaperShip still does a better job than the finally updated official app, but by the time both of these came out I’d long since moved on.

    I like that they developed an API, but it still leaves a lot to be desired. If you have to query a database in another country and download pdfs again when it sits on the same hard disk, that’s just sad. I mean, seriously, it’s much easier to reverse engineer the SQLite database they use, than use the official API.

    Not a fan of having everything in the cloud. I want to be able to choose what stays just on my hard disk and what contributes to the cloud quota. Mendeley has the option to choose this, but when I experimented with it the option wasn’t respected, everything went to the cloud anyway.

    The automatic meta-data extraction is excellent though, and just for this feature it might be worth persisting with it.

    After going off it and returning to it, the thing that put be off finally was that there is no item links. Mendeley desktop just won’t play nicely with others. Most of the other software I’ve tried allows you to copy a special item link that I can paste into other software, clicking the link opens the paper in the software. These links lets you use other software with the reference manager, but not Mendeley.

  • Sente—The continuous updating between different devices is really nice. You can annotate on the iPad and have the annotations appear on the Mac almost immediately. If Sente had had a pen-annotation feature on the iPad I would probably have been a long term fan. When I tried it, full text search was also not available, and it took quite a while before support even for retina screens came out. There is also weird decisions around the implementation of tags which I just found irritating. Having Quick Tags and ordinary tags (which were quicker to enter) differing in function but presented the same in parts of the UI is just silly.
  • Papers—I started using Papers at version 2 and immediately switched to version 3 when it came out. Oh boy. You just need to google “Papers 3” to see a long litany of problems. After trashing my collection for the third time I gave up on reference managers entirely and moved to DevonThink. The development of Papers seems to follow the credo “add gloss first, fix problems later. Much later.” Ironically, I’ve gone back to papers now it’s more stable finally (and had to trash the collection yet again as the old version just refused to update).
  • Zotero—This is the only open-source alternative that I’ve used for any length of time. It’s functional, just not very nice to use when you want it to manage pdfs also. Very early on I used Zotero before switching to Mendeley. Having everything in a browser is not as nice as a native app.
  • DevonThink Pro—This is not a reference manager but more of a general document manager. DevonThink reminded me of the lesson Google brought to everyone—search is really powerful if done right. It was surprising to find that all that meta-data wasn’t that critical. Being able to flexibly organise material and search the full text more than makes up for it. Well almost. It would be really useful sometimes to be able to grab all the articles by a particular author. If DevonThink had arbitrary key-value metadata with each item, and allowed scripting in a decent language like Python (hint hint) it would rock. Unfortunately the iPad app is just a reader so I can’t annotate or organise the material with it.

In general the iPad companion apps are a long way behind general pdf annotation apps that are available.

So, where does that leave me? None of them fit with what I think reference managers should be capable of, thought there are some nice features out there. If you were to combine the sychronisation of Sente with the meta-data extraction of Mendeley, the gloss of Papers, the open-source Zotero, and the search, flexibility and indexing of DevonThink. Acquire a pdf annotation app, and make the whole lot scriptable, I’d be in heaven. I’m currently using a combination of DevonThink and Papers 3 grudgingly as the best of a bad bunch, wincing every time I heavily annotate a paper knowing that I’m probably going to lose those annotation in the future.

Sadly, the long term solution seems to be to write a new reference manager.

© Copyright 2022 Alexei Gilchrist