Re: Review of draft-bberry-pppoe-credit (Informational)

What?  That's not the draft I thought it was.  How stupid can I get?

Having now bothered to read it, I can't help wondering what PPPoE has
to do with the problem or the solution.  Why can't you do the same
things without having wasting precious radio bandwidth on 802.3 headers
and trailers?

I'm confused by the diagram and words in the introduction.
If I take the diagram literally, I wonder:

    Why don't the ends of the RLP link in the diagram in the introduction
    need their own authentication, authorization, etc. controls on
    access to the radio link?  If they do, then why wouldn't it be
    better have 3 separate PPP sessions?  If so, then why is it necessary
    to modify the PPPoE protocol?

    Are there any realistic hope that the vendors of the boxes
    labelled "Access Conc." ship such changes to PPPoE?
    If not, isn't that another reason to have 3 separate links?

    It seems to me that there is a continuing, fundamental confusion about
    network layers here and in the draft I thought was at issue.  The
    confusion is in seeing PPPoE as something above the link layer and
    equivalent to IP.  This and that other draft seem to view PPPoE frames
    as packets that should be shipped half way around the world through
    all manner of gateways, routers, and forwarders.

    This draft talks about "efficient routing" using the "link metric
    mechanism." Is that a wedge to introducing some kind of routing protocol
    into PPPoE?  I don't see much else in that vein in this draft, but it
    sounds like a start toward recapitulating link state vs. distance
    vector etc.  and so forth and so on.

    If instead you have 3 separate link-layer links, 2 PPPoE and one other 
    above or otherwise involving "RLP", you could use things are already
    tolerably well understood.  You'd probalby need to assign IP addresses
    to the Host Radios

But on second thought, it makes no sense that the pair of stationary
access concentrators should be connected via mobile radios.

Contrary to the diagram, is the idea to have new access concentrators
that talk via radios to mobile hosts, with IP over PPPoE over RLP? 
Is the diagram more about layering that links between boxes?
If so, why do you need PPPoE?


Vernon Schryver    vjs@xxxxxxxxxxxx

_______________________________________________
Pppext mailing list
Pppext@xxxxxxxx
https://www1.ietf.org/mailman/listinfo/pppext