You are right, Luke-Jr. In most cases you'd probably want a different address for new transactions.
There are some use cases I was inspired by:
- Some some people (and websites) publish a bitcoin address where you can send donations; it would be much nicer if those are human-readable, instead of some kind of "garbage".
- In online (chat / forum) communication, copy/paste is available, but a human friendly address is much easier to use when communicating over the phone.
Now, taking your comment into account, I propose the following scheme:
- In DNS, publish the location of your mapping server, e.g. a _bitcoin._tcp.example.com SRV record
- This SRV record points to a mapping service, which accepts TCP connections
- The mapping service has a pool of addresses for each user it knows
- As a sender, you can connect to the mapping service, send the request "Give me an address for firstname.lastname@example.org
" and you'll get a bitcoin address returned
- As the owner of the address, you can connect to the mapping service, send the request "Update my address pool. This is the new set of addresses you must use. This message is signed by the private key attached to one of my addresses you currently know of" and the server will update the pool.
This would allow administrators of a domain name to set up the server, set up DNS, for each user manually enter a single address and after that, the system can be used zero-config by senders and also by the user itself to update. The update could be added in the bitcoin clients, such that it automatically updates the address pool when it detects the addresses have been used.
A variant: instead of an SRV record and a TCP service, another kind of record (maybe TXT) could be used that points to the HTTP address of a webservice, the way Luke-Jr uses in his post.