MVP0.2: Collaborative PDF Reader

[strin]

• Collaborative inline annotation
• Comments
• Share through link (this is also a landing page)
  • user can copy and paste paper link (maybe just from browser address bar)
  • new user can view comments through the shared link
  • new user has the option to save the paper
  • saving the paper asks for sign up (thus we can acquire users through sharing)
• Instant load
  • pdf2html
  • optimize load time of html format paper (first render low-quality first page, and then the entire paper)

[larryxiao]

Design for Comments (Draft)

Building a comment component.

Background

Currently we have pdf (paper) converted to html using pdf2htmlEX. It parses the pdf file and generates css, html and javascript. The html is the pdf decomposed into the pieces of text or pictures it has. The css records all possible style related info about the pieces (font, line spacing, relative position to page etc.). The pieces are labeled using the styles from css, so they are unique. The pdf provides animation for scrolling and other functionalities.

Design

Options

• Use external libraries / tools
  • Side Comments - Medium.com style commenting provides the frontend for comments.
    • We need to add each of the div in our html of its class, and a unique id. This requires either one extra pass annotating the html, or do it dynamically and record info about the change in database.
    • The representation is decentralized, and require user action for discovery.
• Implement by ourselves

Backend

Use redisdb for storage, data model:

threads {
  paperid
  threadid, used to locate all posts
  meta data {
    pos, position in paper (can be used for on demand loading)
    npost, number of posts (can be used for preview)
  }
}

comments {
  threadid
  commentid
  usedid
  text
  datetime
}

Frontend

The comments section is a side pane, it can be toggled by user.

When rendering, we give a hint on the pdf indicating it has comments somewhere, user can click/mouseover and open/see preview of it. When the side pane is toggled, we also indicate how threads correspond to original texts (simultaneous highlight or a line connecting the two).

For commenting, user can add reply to existing threads. Or click on the pdf to create a new thread. We can use a global listener, and identify the element clicked jquery.event.target, use the xpath or css selector as ID for thread.

Discussions

• Managing thread is extra work.
• The processing of pdf is risky (security),
• The html representation of pdf is not elegant >..<, is there better way to do it? I checked pdf.js by Firefox, and mendeley web, the pdf is presented in similar way (text + style info). So I imagine there will be much dirty work involved in frontend implementation.
• I think Mendeley should be implementing some collaboration functionalities, because they have a almost full fledged web app.

[strin]

Cool. Awesome design. I say let's just integrate medium.js and see how it works.

• I think pdf.js and mendeley also renders pdf to html. The question is how to make user experience better.
• For most papers, I thought there aren't any encryptions, no? Could you explain what the security issue might be?
• Mendeley could be doing this. But not many people I know use Mendeley, because it's too complex....

[larryxiao]

Ok let me try to make a fast prototype.

• Yep, so my opinion is that it require some good effort to make user comfortable.
• That's for the security of the server that we don't need to care about so much for now. I think I saw Imagemagic had some security bug.
• Looping back to the first point, we need to make a usable comment prototype fast because there are many similar paper reading software that can be working on this. And I feel we might do better with some high barrier stuff, such as meeting transcription, with our resources.

[strin]

Cool. I kind of want to keep our effort minimal. Some higher barrier stuff is more interesting. Let's wrap this up and build something for our own daily uses.