News for hackers. Stuff that matters.
Re:PDFy – Instant PDF Host by walterbell on Tuesday July 15, 2014 @01:44AM (#8034977)
Ideally, metadata extraction would be done on upload and presented to the user for optional manual correction. This would be a major contribution to findability, because PDFs often have incorrect metadata (try searching for anything on archive.org) or the person uploading may have metadata relevant to a use case that is unforeseen by the document creator.

In fact, the same file could be uploaded multiple times with different metadata. There's room here for experimentation, e.g. publishing content hashes and linking "duplicates" that have different metadata.

If PDF metadata can be published in a structured format, it should be then be possible for Calibre or Docear / Zotero to import the PDF + JSON metadata directly into the document database.

Re:PDFy – Instant PDF Host by joepie91_ on Tuesday July 15, 2014 @01:50AM (#8034987)
The problem is that I don't want to add any more roadblocks to the upload process - it should be as 'instant' as possible. Even the 'public' vs. 'unlisted' selection took quite some consideration.

Perhaps crowdsourcing metadata might be an idea - but that involves quite a bit more complexity, implementation-wise.

Re:PDFy – Instant PDF Host by walterbell on Tuesday July 15, 2014 @01:53AM (#8034994)
How about an "edit metadata" button linked to the submitter's session cookie, which is only active for a few minutes, similar to HN post editing? For those who don't care, upload process stays the same. Those who want to edit have the option, within a few mins.
Re:PDFy – Instant PDF Host by ratpik on Tuesday July 15, 2014 @03:35AM (#8035238)
Why not build a re-CAPTCHA type service around crowd sourcing PDF metadata. Read a PDF while you wait for something to happen.

Too idealistic?

Re:PDFy – Instant PDF Host by fiatjaf on Tuesday July 15, 2014 @07:50AM (#8035847)
Linking hashes would free a lot of space, and letting each user place its own piece of metadata on each hash it wants would solve all the metadata and user trustworthiness problems.

Google would do the rest.