Launchpad Entry: Integrating with Ubuntu One
Packages affected: python-contacts-api, dbus-contacts-api, gtk-contacts-picker, erlang, couchdb, couchdb-bin, ubuntu-couchdb, couchdb-oauth, couchdb-history, ubutuone-sync-ui, evolution-couchdb, akonadi-couchdb, tomboy-couchdb, firefox-bookmarks-couchdb, python-couchdb, couchdb-glib, dbus-couchdb, python-couchdb-records-api, dbus-couchdb-records-api, icontool, ubuntuone-client, ubuntuone-storage-protocol, python-oauth, ubuntuone-screensharing, ejabberd, configglue, gwibber-couchdb
Ubuntu One provides capabilities for the Ubuntu desktop as well as extended capabilities via optional online services. Integrating Ubuntu One capabilities with the Ubuntu desktop will provide storage sharing & synchronization (across a LAN and (optionally) the Ubuntu One cloud), screen sharing, and data (contacts, notes, Firefox bookmarks) sharing & synchronization via CouchDB.
File Sharing & Synchronization
File sharing and syncronization allows users to share and synchronize files across their local network. An optional capability will allow users to easily share and synchronize files across the Internet using the Ubuntu One online service.
The Ubuntu One client will no longer use a client applet (as is the case with the current beta). Instead, the client will use downloads, FUSA, and standard Ubuntu messaging indicators.
Provides screen sharing between two desktops using Empathy, Telepathy. The "Share My Desktop" dialog will be unaffected. An optional Ubuntu One service will provide a proxy (utilizing ejabberd on the server) to allow users to more easily share their screens across the Internet.
Data Sharing & Synchronization
Overview: Apache CouchDB is a distributed, fault-tolerant and schema-free document-oriented database accessible via a RESTful HTTP/JSON API. CouchDB is the database for Ubuntu One data sharing and synchronization capabilities.
Erlang is the language and runtime environment for CouchDB. Work will be done to trim the Erlang package so that it is as small as possible to run CouchDB and other applications dependent on the Erlang runtime.
Each user will run their own instance of CouchDB, which is different than the standard install of the current CouchDB package. There is also a need to setup various CouchDB configuration files, design docs, and triggers. Additional capabilities will be added to CouchDB, including OAuth support and document history.
In order for developers to more easily use CouchDB in their applications, we will provide libraries for a variety of languages and environments, including Python, DBUS, and GTK (GLib).
A new desktop application will be developed to allow users to easily share and synchronize their data (contacts, notes, bookmarks, etc.) stored in CouchDB between Ubuntu computers on their local network.
Providing a CouchDB contacts database that all Ubuntu applications can use is one of the key applications of the Ubuntu One data sharing and synchronization capability. Integration with Evolution, Akonadi, and Gwibber will store contacts in CouchDB. Developers of GTK applications will have a GTK widget which can provide their applications with a way to easily lookup/select contacts from the contacts CouchDB database. There will also be a contacts API for Python and DBUS, which will allow developers to use the CouchDB contacts database without worrying about the underlying details of CouchDB and the contacts schema.
A Firefox bookmarks synchronization and sharing extension will store Firefox bookmarks in CouchDB, which will provide sharing and synchronization capabilities.
Tomboy will use CouchDB for storing its data which will provide sharing and synchronization capabilities.
The packages below include the Python OAuth library and a tool for managing Ubuntu One configurations.
Specific code changes will be captured here. The summary given above in the design view highlights the major changes overall. As key code changes in the packages listed above are identified, they will be listed in this section.
This should highlight any issues that should be addressed in further specifications, and not problems with the specification itself; since any specification with problems cannot be approved.
BoF agenda and discussion
- ubuntu-one uses the couchDB database
- couchDB supports replication, it's very simple, open and uses structured data
- ubuntuone provides an API to couchDB
- couchDB has good upstream support, and we need to get it into main for karmic
- the service in front of couch supports http, which simplifies access for the API
- lazr.restful, a python api originally developed for launchpad
- couchDB needs erlang
- erlang is about 45M installed size
- maybe we can strip down erlang to make it fit more easily fit the single live CD
- the application only needs to know how to handle the structured data with couchDB, not worry about things like sync, replication, etc
- one use case, having a syncronized addressbook
- evolution-data-server directly queries ubuntuone (locally)
- the contact database from ubuntuone does sync locally, so e-d-s doesn't directly query the server just the local sync of the service
- for the contact storage, what is the upgrade path?
- start time is something to consider, the syncdaemon runs continuously, but it is in the background
- perhaps defer the rescan to speed up login time
- working with the design team to work out the details of what to do with the applet
- it might go away when idle
- other uses:
- screen sharing
- for now sticking with current technology, vino and vnc
- bookmark syncing
- trademark issues with customizing firefox
- perhaps weave integration
- modifications would be made in ubufox
- tomboy sync
- sync notes to u1 instead of snowy
- should we consider gnotes?
- backing up a usb drive when plugged in
- screen sharing
- What KDE apps should integration with u1?
- system wide storage called akonadi using mysql
- ultimately all configuration should be done in the applications, but we might use a common u1 control panel to configure various aspects of u1 until we get that level of integration
- we need to work on the MIR for the u1 client and deps (including erlang) for karmic
- before doing the MIR we should figure out how we can split erlang into smaller chunks for CD size constraints. See: