- From: Rich Tibbett <rich.tibbett@gmail.com>
- Date: Wed, 23 Jun 2010 10:47:56 +0100
- To: public-device-apis@w3.org
- Message-ID: <AANLkTinnbWQzWjx4mnwPAHReSTSxnSltmuDdMrNqTCyt@mail.gmail.com>
I'd like to request that we drop the serviceId attribute from the Contacts API specification. Following is a description of why this is important, if not essential for users and their data. As it is currently provided, the serviceId attribute cannot reasonably identify a contact record that has been sourced from a plurality of address book sources. For example, if a contact record is the result of combining data from a user's device-based address book and Facebook/LinkedIn/Plaxo, the serviceId is entirely incapable of reasonably expressing the source of this data in the serviceId field. Removing this attribute removes the ability for web apps to change the storage location of a particular contact record. For example, an operator-driven web application may attempt to move all contact objects from cloud/networked sources to the device-based address book - possibly based on some (opaque?) policy rules. To do this the web application would need to identify which records are not currently stored in the device, and then edit them appropriately. Equally, a non-operator-driven web application may attempt to move all contact objects stored in the device to the cloud in the same manner. Importing and exporting between address books should not be an intended use case of the Contacts API exposed to web applications. Importing and exporting between address books should be left to other interfaces at an implementation's discretion. Removing the serviceId attribute is the best way to ensure that the current back-and-forth 'who owns the data' battle doesn't spread to the web platform itself while still letting implementations themselves decide the most appropriate data stores to use. In cases where an implementation is required to maintain sync with an actual data source, this should be accomplished via an implementation's configuration panel. The implementation itself may maintain the actual data source for individual ContactProperties as required, and sync as appropriate (e.g. [1]). When properties are changed or added, the implementation should sync with the necessary services as appropriate and based on the capabilities of specific data sources to store any ContactProperties being set. This back-end interaction is entirely obfuscated from web apps. Notably, serviceId is the only non Portable Contacts property included in the entire specification. Removing this attribute makes the W3C Contact schema a direct 1:1 mapping of the Portable Contacts schema. Please let me know your thoughts. Incidentally, for an example implementation of this proposal please see the excellent work of Mozilla [1] [2] [3]. This discussion relates directly to ISSUE-43. regards, Richard [1] http://www.open-mike.org/entry/contacts-in-the-browser-0.2(source-attributed data as it is stored in the Mozilla Contacts implementation - with links maintained to each address book provider as appropriate) [2] http://mozillalabs.com/files/2010/03/servicesView.png (Mozilla Contacts import functionality screenshot for merging data from a plurality of sources) [3] http://mozillalabs.com/files/2010/03/contactView.png (Contact data as it is exposed to web apps by the Mozilla Contacts API (notably, minus 'serviceId' location))
Received on Wednesday, 23 June 2010 09:48:49 UTC