netdev
[Top] [All Lists]

Re: ppp control frame passing (was: (none) / Re: your mail)

To: ak@xxxxxx
Subject: Re: ppp control frame passing (was: (none) / Re: your mail)
From: Henner Eisen <eis@xxxxxxxxxxxxx>
Date: Sat, 18 Mar 2000 16:56:09 +0100
Cc: ak@xxxxxx, kai@xxxxxxxxxxxxxxxxxxxxxxxx, netdev@xxxxxxxxxxx, i4ldeveloper@xxxxxxxxxxxxxxxxxxxxxx
In-reply-to: <20000318131637.A5599@xxxxxxxxxxx> (message from Andi Kleen on Sat, 18 Mar 2000 13:16:37 +0100)
References: <Pine.LNX.4.10.10003160943150.17180-100000@xxxxxxxxxxxx <20000316123233.A1849@xxxxxxxxxxx> <ouvh2kfyti.fsf_-_@xxxxxxxxxxxxx> <20000318131637.A5599@xxxxxxxxxxx>
Sender: owner-netdev@xxxxxxxxxxx
>>>>> "Andi" == Andi Kleen <ak@xxxxxx> writes:

    Andi> It is probably easier to use a special scheduler for that
    Andi> that calls the normal scheduler as a child (and make sure
    Andi> PPP devices always have that special scheduler pushed). That
    Andi> scheduler would make sure that control packets come always
    Andi> first.

Sounds good! I think we should take that direction. This would also be totally
independent of the underlaying ppp implentation. That means, if it works,
it can be re-used for the other ppp implementations as well (syncppp, generic).
Then, pppd support can migrate to using packet sockets instead of /dev/*ppp*
and all ppp stacks can migrate smoothly. On the long run, I think we should
support the generic ppp by isdn, but with the current call / demand
dialing this won't be easy befor a major rewrite of the isdn link level
has been done. (This rewrite would of corse use the generic ppp).


Henner

<Prev in Thread] Current Thread [Next in Thread>