Date: Fri, 27 Oct 2000 21:24:08 -0400 From: "Louis A. Mamakos" <louie@TransSys.COM> To: gcorcoran@lucent.com Cc: Julian Elischer <julian@elischer.org>, freebsd-net@FreeBSD.ORG Subject: Re: More on PPPoE & ADSL (Telstra Bigpond) Message-ID: <200010280124.e9S1O8G52999@whizzo.transsys.com> In-Reply-To: Your message of "Fri, 27 Oct 2000 18:56:28 EDT." <39FA081C.3E56D791@lucent.com> References: <F50iFEW6sStwNeKjUbE00001146@hotmail.com> <39F8C29F.D785C588@lucent.com> <39F9210E.B728D4F8@elischer.org> <39F9B679.CA563B9E@lucent.com> <39F9E669.FB8D77D2@elischer.org> <39F9F1FB.F00E686F@lucent.com> <39F9FFAD.2992767D@elischer.org> <39FA0056.8CB7D452@elischer.org> <39FA081C.3E56D791@lucent.com>
next in thread | previous in thread | raw e-mail | index | archive | help
> Personally (my opinion only), I dislike PPPoE. It adds a full THIRTY (30) > bytes of overhead to every packet you send! And for a (real) DSL link, > it's not needed. That is, you really just want to send PPP over ATM > (DSL packets get formatted as ATM cells, if you didn't know). The 30 bytes of overhead on the link aren't that big a deal considering that it's all being shredded into AAL5 ATM cells (usually) too. The speed of the DSL span likely isn't the limiting factor; it's the oversubscription beyond the DSLAM. While you can certainly do PPP over ATM, this also means you get to buy a *DSL NIC card for your PC, and figure out how to write a driver for it under FreeBSD. The alternative that PPPoE gives you is using a $15 10/100 Ethernet NIC that's already supported, plus you can have multiple end systems beyond the *DSL CPE modem which can simultanously use the link. The PPP over ATM alterative means that the system with the ATM NIC needs to act as a router, and the overall reliability is only as good at the one system. While this isn't a big problem for FreeBSD, consider the household with only Windows boxes, and the frequency at which these things are restarted. When we started doing some DSL development work at UUNET, some of the explicit goals were to: - support multiple sessions on one DSL connection, potentially different end-systems - to enable immediate DSL deployment without requiring new DSL CPE hardware to be built. At that time, pretty much every vendor had dumb ethernet bridges for their particular flavor of DSL. Note that some DSL implementations are not ATM based - CHEAP, and hopefully already installed, network interfaces for the end systems - as little configuration of the DSL CPE device as possible. No config at all is great. For a residential service, controlling customer service costs is very important. A somewhat unstated goal was to make it easy to add DSL support to other than just Windows platforms. There are a bunch of alternatives which are based on the DSL/ATM NIC card in the client end-system; these suffer the system reliabilty I mentioned. But as a FreeBSD user, it seemed pretty clear to me that there was not going to be much motivation for the folks building that hardware to release documentation, much less support "non-mainstream" driver development. By putting all of the DSL hardware behind a cheap and ubiquitous Ethernet host interface, most of those considerations go away. There were other proposals using Ethernet; one particularlly scaring one had ATM cells tranported over the Ethernet to the end-system, where the SAR (cell segmentation and reassembly process) would BE DONE IN SOFTWARE, along with a complete ATM signalling stack! This makes sense if you consider that Intel was proposing it, and you're looking for more opportunities for selling CPU upgrades. Now, if all the f*&king firewalls in the world didn't break MTU discovery, using PPPoE would be much more painless, sigh. louie To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-net" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200010280124.e9S1O8G52999>