[Monotone-devel] using monotone for a simple logging database

[John Wright]

I am looking for some architecture direction on how I can leverage some of the netsync code from Monotone in a project I am working on.  I am creating a unique private logging app on top of sqlite3.  By logging I mean  a simple diary where users will log small micro-bits, ala Twitter or Jaiku and I will pull a user's public entries from Twitter and other sources in to provide context for this micro-journal.   I have been trying to design an architeture that combines feature of a dvcs like monotone with sqlite because:

• I want a portable single file database that plays well with in my different application environements like the web, the Iphone, and Ruby on Rails.  So I am developing on top of a simple logging schema in sqlite.
• I want it to work equally well in offline mode and to easily sync when reconnected.  I liked the git push and pull work and dream of pointing to a url of another sqlite database and sync the entries between them over http.
  
• Ala git, it should be easy to deterimine if an entry has been tampered by comparing with a SHA1 hash of the entry to another instance of the logging repository.

Unlike a DVCS my schema is flat, there is no notion of a tree of edits/commits, just a running "straight" history of logs.  Actually there are no file blobs, just the log part of a commit.  I am hoping that this can means I can implement  this dvcs -lite logging repository in _javascript_, Objective C, and Ruby without having to resort to anything but SHA1, basic language primitives and http and things available in most programming languages.   I am writing you Monotone guys hoping somebody can help point me the relevant bits of Monotone that I can look at for inspiriation.  I am of course digging around myelf but somebody might be able to expedite the process for me.  I am particurlarly interested if whether some of the netsync code can be borrowed for such a purpose.   I know that in netsync, partial hashing of the sqlite file is used to compute what needs to be sent between the 2 systems.  Is it possible to implement a simplified version of this protocol that could work even with a _javascript_ client as one of the peers.

If this isn't an appropriate question for this forum, please ignore and my apologies,

Thanks!

John