VanLUG Email Archive
> R Garth Wood writes:
>
> > I looked through that rfc but it had no rational included.
> > Can someone convince me that settting your dns
> > with pppd is a good thing(I know it's
> > a usefull thing to some ppl)?
>
> >From an operations point of view it's great. If an ISP changes the
> address of a name server then it's automatically propagated to those
> customers using the feature, possibly cutting down phone calls
> considerably.
Why don't we use ppp to configure the whole machine then?
We'll add partitioning and formating, hell let's make
a ppp os! I know it's USEFUL but so is logging in as
root all the time. My example is hyperbolic, I know,
but it underscores my point. My proposal would
be to run an equivalent of ip-up with a prog
that queries the other end of the ptp link
to find the name server. that way OSI is
still happy(err kind of).
> > I don't believe setting up application lvl software
> > belonghs in the datalink/network layer(s).
>
> Pppd is application software that sets up and maintains the link
It does more than just the link lvl it sets up routes as well.
> level. Why shouldn't it start up an application level script or edit
> a config file itself based on information it's negotiated with the
> other end?
It should; but JUST the link, not other ones.
It should send the eqiv of an interrupt saying link lvl's done.
> My opinion is slightly biased against it doing so simply because I've
> always believed resolv.conf is a file for humans, but OTOH it can be
"file for humans"? Explain.
I don't know what OTOH stands for. :<
> thought of as a cool hack to automatically maintain resolver settings.
It's a hack...
> One could argue that RFC 1877 is incomplete because it only addresses
> name server IP addresses. What about all the other info needed to
> configure a resolver?
See above. Not a bad idea, though. My method transends ppp
and could be used for broadcast networks, too. I guess
it would be similar to bootp(although I've never used it).
> > And yes I'm not too fond of defaultroute either.
>
> Without a default route your routing table would be *very* large.
> There're about 50k routes on the net these days.
Ha! I'm not proposing to bring back HOSTS.TXT! ip-up
could perform the same function as the option defaultroute,
that's what I meant. And again application lvl software
could be notified when the ppp link is up and make a
better choice for the default route.
Example: suppose you have many routes but they are unrelyable.
The interface may still be up but a link downstream might be toast.
Fire up a ppp link to isp1. If that is down go through a list
of other links.
+-----------------------------------------------------------------------+
| R Garth Wood | <insert witty comment here> |
| | -R G Wood |
|
| |