I'm happy about all movement in this space, but I don't really see the unique value proposition yet. What is special about papis?
Being hackable (python) is certainly nice, but many of the steps listed on the homepage are a lot faster and easier with other tools such as Zotero or straight BibTeX. Especially manually entering titles or other metadata on the cli - that would be the seventh level of hell for me. I want one click on a paper's webpage and be done.
You don't normally need to enter metadata manually. You can import it using the publication's doi, arxiv etc. Then papis gives you an opportunity to review and update the metadata. The other functionality that papis provides is a quick search through titles, authors etc.
Of course much of that can be done with doi2bib [0], arxiv2bib [1], etc., which you can combine with the wonderful bibtool [2] to keep a clean bibtex file. That's what I use and the advantage of this over papis is that you can version control it.
That said papis has its use. It's less heavy than zotero, a bit less proprietary format (the info is stored in yaml files iirc) and it provides a layer over bibtex. If it had full text indexing and search I could see myself using it more.
Yes, many workflows are already made up of a lot of tools that are individually great but require a lot of duct tape. I guess the best path for a new effort is to work towards making any aspect of it more automatic. That's why even entering a DOI by hand seems like a regression, when Zotero can extract it automatically in most cases (as well as ISBN for books etc.).
Some areas where I would love to see innovation, and that seem entirely plausible for papis:
- Collaboratively clean metadata (i.e. share sets of papers with collaborators in a less silo-esque way than Zotero's online libraries
- Do a better job at collating related versions (i.e. multiple related talks, preprint, online first and published version) ...
- Automatically track and import new papers into staging area
- Automatically fetch PDFs from some external source (and share with collaborators)
Most of these are possible in some existing apps, but it seems that the designers always expect people to joyfully curate every nook and cranny of their libraries, when in reality I have five duplicate versions of many papers, each missing some crucial metadata, none being linked to the one PDF file that's somehow stuck on a WebDav share, and Google scholar metadata polluting BibTeX fields but omitting crucial details such as the publisher.
This field really is another case of "we can re-land and re-use rockets, but digitally organizing a theoretically well-structured information space is impossible".
I agree. Collaborative editing in this space is awful.
> This field really is another case of "we can re-land and re-use rockets, but digitally organizing a theoretically well-structured information space is impossible".
Maybe because it doesn't matter that much to most people? People who live in uncommented F77 codebases are lightyears away from worrying about properly archiving information.
"Use zotero conectors
Just install the zotero connector browser plugin here and type
papis zotero serve
to start listening to your browser for incoming data. Whenever you click the zotero button to add a paper, papis will add this paper to the library."
Being hackable (python) is certainly nice, but many of the steps listed on the homepage are a lot faster and easier with other tools such as Zotero or straight BibTeX. Especially manually entering titles or other metadata on the cli - that would be the seventh level of hell for me. I want one click on a paper's webpage and be done.