There are clearly many areas that could take advantage of easier networking, however there are two that really stand out - incorporating service location into widely used servers and clients, and multicast DNS.
A major planned activity is to develop changes against glibc to add a multicast DNS capability. This is effectively against the Bind 8.x resolver, which should be widely portable to other systems.
There are a number of aspects to this, including integration into a typical name server switch (NSS) style design. Among the design issues for this are how to handle the various name spaces, since it is generally accepted that the multicast DNS namespace (or Link-Local Multicast Name Resolution) is not the same namespace as the DNS namespace(s), but there rules for doing a lookup from the application are not well understood.
Development of a multicast DNS responder may also be required.
As of writing, Apple had just released their new web-browser (marketing name "Safari"), which can use DNS service discovery to locate web-servers. Adding this type of service advertisement capability to various servers, and suitable service discovery capabilities to major servers is clearly important for the on-going acceptance of open source software.
It would be very convenient to have a single API that allowed service location using whatever protocols are available, perhaps using some form of failover system like that currently provided by the Name Service Switch.
This would require adding DNS Service Discovery to any multicast DNS responder, and possibly creating a new API for service registration and location.
| <<< Previous | Home | |
| Work to Date |