root/feedmelinks/import/Docs.txt

rationale
======================================================

Importing bookmarks (or favorites) is important because users can
preserve their existing data -- what good is a solution that forces
users to throw away all their work? No way, man!


implementation plan
======================================================

0. user arrives at import screen, sees a picture of the 3 step process:
   "upload, confirm, import!". user uploads the bookmarks file.

1. we parse the bookmarks file, ignoring any unparseables, and we
   generate two "prep files": LINKS.RAW, a file of url / name / taglist rows,
   and TAGS.RAW, a de-duped list of all tags in the file.  (we do this
   conversion because the raw bookmark file is a pain to iterate over
   and we'll need to traverse through this input data more than once.)

        > log: parsed N links from U's bookmarks

2. bounds-check the number of link and tag insertions (line-count
         of LINKS.RAW & TAGS.RAW) to detect attacks from D.O.S. dorks

3. randomly sample (both contiguous run and scatter) the links to
   check for uniqueness against other links in the imported bookmarks
   file as well as uniqueness against the user's existing FML links
   (to detect possible link spammers)

        > log: any possible violations from 2, 3

[if OK to proceed, i.e. no grievous errors...]

4. now we exhaustively loop over LINKS.RAW(looking at just the
         links) and TAGS.RAW against the user's links in the FML
         database, and stream the files back out to disk, marking
         an X before any link or tag found in the DB. Save the new
         files as LINKS.READY and TAGS.READY

 > log: number of import dupes for U

5. display the links to be imported by looping over LINKS.READY
         (with a link & tag count), and showing all the links with any
         duplicates in gray, and let the user OK the import. (remind
         them that all these links will be private by default.)

6. using the prep files, generate the proper SQL commands to INSERT
         link, tag, and link<->tag rows and run it. ideally
         double-check the generated SQL for any glaring security
         (SQL injection) bugs (might be good to run the SQL on a
         temp table and look for errors, although that's no guarantee)

        > log: any SQL anomolies as possible d.o.s. dingus attempt by user U with code C

7. remove any temp/prep files

8. teach user how to manage the new, imported links; design a nice
   screen to show the user how to make their imported links public
   and re-tag as necessary






uncaught weaknesses: vulnerable to link-name spamming until we check those for dupes too

vulnerable to query string spam






a few words about the netscape bookmarks file format
======================================================

when you see <A HREF="ANYTHING_BUT_UNESCAPED_DOUBLE_QUOTE"
        this is a link.
        if there is a currently open link, close it and add it.
        
when you see a <DD>STUFF^WHITESPACE<DT>
        this is a description, grab "STUFF" and add it as a comment the current link
        this link is now done. add it.

when you see <H3 ADD_DATE=
        this is a new folder. add it.

future
======================================================
how do you parse apple's .plist files?