Large DSL networks are changing to PPPoE. This will bring back the username/password login/logoff concepts to DSL connected PCs. They are doing this by switching to PPPoE.
PPPoE, Point to Point Protocol over Ethernet, (aka RFC 2516, and by the way, if you think RFCs are boring, try 1776 or 748 to name but two), is a method of connecting you to your ISP. Based on an existing standard (PPP), which was developed for encapsulating (carrying) multiple protcols over point to point links, PPPoE won a small battle with a microsoft backed solution to become the leading contender.
The main characteristic of PPPoE is the two concepts of authentication, and session.
Why
PPPoE is one solution for coping with a few management problems inherant in typical DSL and cable networks. Bell Canada Sympatico is scheduling a switch to PPPoE this month, and Bell Atlantic was going to switch in the DC area last week, but called it off due to client software hassles (more on that below!). UUNET/MCI/WORLDCOM/SPRINT whatever it is called now is also a big PPPoE adopter with its DSL offerings.
What was the problem before PPPoE?
1. Broadcast traffic. non-PPPoE DSL implementations allocates each machine an IP address, whether statically, or by DHCP (dynamically). Broadcast IP traffic is meant for all addresses within your "subnet".. this has advantages in local area networks: by "calling out" for service or "broadcasting" information, a common class of network tasks becomes simpler. In a DSL network, however, it is not desirable, either for efficiency or for security, for random groups of unrelated users (that happen to fall in the same subnet) to see broadcast traffic from other users. For example, if someone runs appletalk on a hub with a regular DSL modem, a lot of meaningless traffic will go out to anyone else with nearby IP addresses. Windows netbios also favours broadcasts, (to its disadvantage).
2. IP and session management. When a large telco DSL or cable network is required to give subscribers freedom of service provider choice, they many have network, IP and routing management problems. Enforcing PPPoE on the user side effectively creates tunnels between the user and their chosen service provider, insulating the carrier from the traffic, and simplifying the relationship between user and ISP. In addition, session management gives the internet service providers the ability to more closely monitor usage, via the authentication mechanisms that come with PPPoE, in a way compatible with their old dialup authentication and accounting systems.

DSL as a TV set? After writing this piece it was pointed out to me that one advantage of PPPoE is that it allows for multiple services. These are shown as being potentially available (by name) to the PPPoE user before authentication. It is promised that these services might be expanded from the current number of one (1), the internet, to include others.. premium speed downloads? pay per use? who knows.. it sounds to me like the services concept was added because it was possible, rather than to fill a need, leaving someone to find uses for it later. Personally I'd rather leave such service or channel selection technology at an application layer (isnt that what the web is all about?) but possibly there is some subtle networking advantage to be had offering service selection at a low level.

What does it mean for you?
From the DSL user perspective, PPPoE is actually a step backwards, at least in terms of setup and complexity, over current DSL implementations. To enable a computer for PPPoE, software must be installed that handles the session setup and authentication. DSL providers are likely to look at anything other than a windows/mac machine and say you are on your own finding installing this software.
Under windows, this software sits behind the good old dial-up-networking type of interface.. the PPPoE networks will market this as a positive thing "preserving the end-user experience" as they transition from dial-up to DSL.. but in my opinion, it is an experience we could do without. Getting an always-on connection with a public IP address was supposed to banish the horrors of dialup networking control panel and usernames/passwords forever. No such luck.
Secondly, since the PPPoE RFC was only really final in February 1999, software PPPoE packages, vital for rolling out PPPoE across an existing DSL network, are few and unreliable. User experiences with products from www.nts.com (as required by Sympatico), have found that whilst the windows 98 version works, the NT version was a very immature product. The competing software comes from www.ivasion.com, (WinPOET), Linux users must take their chances with a www.nts.com version, also buggy, or gurus can try a pppoe from SuSE. (update: OS/2 does now have the the InJoy firewall), FreeBSD? Solaris? sorry nothing yet [sep 1999]. Whether any of the early solutions out there work well on your particular PC is somewhat of a lottery, and dont expect any real help from the helpless-desk of your huge DSL provider!
mind control
Session authentication is another unknown quantity... requiring sign-on names and passwords mean, for example, two or more computers in a household active at the same time may require purchase of an additional "user id".. to escape that, home or SOHO users may have to share passwords with each other, or NAT type solutions will still be required. PPPoE also encourages the spread of short lease dynamic IP .. discourgaging further the development of home servers at known locations on the net operating over DSL, despite the fact a basically infinite IP space (IPv6) is just around the corner.

WinPOET claims that their software PPPoE setup is just as easy to setup as configuring straight windows ethernet (install it, reboot!), and that may be the plan.. but who amongst you, given a setup.exe, would bet your own money on the total success of a single install on a randomly selected home computer.. I suspect before placing the money on the table, you would be digging around on the disk to check space, running chkdsk, and verifying windows and its DLLS were of the appropriate version.

Hardware to the rescue
Stepping into at least the software install gap, the CPE (customer premise equipment) makers are trying to provide PPPoE solutions in-a-box. One of the first is the new version Nexland internet sharing box.. by building in PPPoE, it stops the PPPoE complexities (incompatibilities, upgrades, hassles) from reaching other equipment on your LAN, and shares the internet with your net. Compatible with any DSL CPE, since it is ethernet to ethernet, it is worth a close look to avoid PPP software hassle. Interestingly, a Nextland equipped network need not change one jot, as a provider switches you from static bridged IP to PPPoE.. (you could take this as a commentary on just whose time is getting saved by switching to PPPoE?).
Conclusion
PPPoE is being rolled out with unseemly haste.. possibly because the larger DSL networks are finding their DHCP servers are not scaled right, possibly because ISPs are complaining they wish to integrate their system accounting for DSL customers in with dial-up, and need the username/password hooks to do that, possibly because Redback is the industry leader and is backing it, possibly because it sets up a framework of control that the large DSL networks believe they will need.. whatever the reasons, if the end-user implementation continues to concentrate 70% on windows 98, 30% on Mac and the rest (NT, Linux, solaris and the rest) can go hang, then it's a bad sign for diversity in tomorrows public broadband network.