The Drupal 8 Multilingual Initiative (D8MI) is aimed at making language support more advanced and widespread both in built-in core functionality and by providing (better) APIs for contributed modules to build in language support themselves. The main problems attacked are (a) insufficient and inconsistent base APIs for language (b) missing language support on some entity types and configuration (c) missing automation for interface translation. The overall architecture plan looks like the following:

We don't start from scratch of course, so we need to consider what is there in Drupal 7 and where do we want to get from there to Drupal 8 by keeping user data intact and gradually migrating user settings and data to the new system as we go. The Drupal 7 functionality is somewhat chaotic and spans many core modules (yellow), contrib modules (orange) and even Drupal distributions (blue):

The Drupal 8 plan is to cover this functionality need introducing most features from the above mentioned modules in core well distributed in an easy to understand fashion. Breaking down to modules, language.module would provide base language handling and negotiation. Modules themselves would do language assignment for nodes, comments, etc. Then the interface translation code itself will keep on as its own service for all modules to rely on:

More details of this breakdown are documented on the subpages and on the issues themselves. Further resources about the initiative:
See all these issues also at http://drupal.org/project/issues/search/drupal?issue_tags=D8MI. This view of issues is automatically cached and might be out of date up to two hours at times.
See all these issues also at http://drupal.org/project/issues/search/drupal?issue_tags=D8MI. This view of issues is automatically cached and might be out of date up to two hours at times.
Based on participation in D8MI issues (font size) using a quasi-logarithmic scale and on recent activity (boldness). Excludes the initiative lead, Gábor Hojtsy.