MultiProtocol per interface support (was Re: LFS-6.0 print process)

Bryan Kadzban bryan at
Fri Jul 9 04:09:18 PDT 2004

Nathan Coulson wrote:
> When I proposed that to some people on IRC, they said it was
> unnecessary bloat...

Bloating what exactly?  It doesn't bloat the ifup/down scripts more than
a few lines (which you're going to have to do regardless of how you get
multiple protocols per interface).  Service scripts may have to be
changed, but they won't get much bigger.  The only difference would be
that they would source $3 if that's set, instead of ifconfig.$1.

The normal case (only one protocol or address per interface) will have
one directory that the old sysconfig directory didn't, but the actual
file in that directory will be exactly the same.  Though in that case,
you might as well scrap the directory and just use a file, which means
nothing is any bigger.  Adding another file won't take any more space
than adding another set of lines to an existing file (except for disk
block slack space, but the same thing happens with init scripts and most
of the rest of /etc already).

I would say there's actually LESS bloat of the scripts themselves with a
directory based approach (compared to the other approaches) -- ifup/down
are about a third the complexity of the SERVICE="static,pppoe" approach,
and about a tenth the complexity of the grub config file approach.  At
least the way I did them.

In short, have they seen the parsing code required in ifup/down for the
other options?  ;-)

More information about the lfs-dev mailing list