[PATCH 1/2] udhcpd: sanitize invalid hostnames to match rfcs

[Isaac Dunham]

On Tue, Oct 20, 2015 at 11:51:51AM +0200, Bernhard Reutner-Fischer wrote:
> [Isaac wrote:]
> >udhcpc and dnsd might be able to use it.
> >Other networking tools (that do DNS lookups) could use it, or could
> >simply trust that if the user inputs an invalid name, either the DNS
> >system will catch it or it will resolve despite the standards.
> >
<snip>
> >
> >> In a first step it would be sufficient to move this code into a
> >function
> >> and then look for more uses.
> >
> >Due to the subtle variations in what we need to cover, a function
> >*will*
> >be larger than the inline code, and will also require thought as to
> >what
> >it should cover.
> >I don't think it's sensible to move it into a function before
> >determining
> >what it should do, and that it really is useful.
> 
> bugs.busybox.net/3979#c2
> 
> I wouldn't support broken names, it doesn't make real sense.
> 
Agreed.

What I'm not sure about is what response to bad names does make sense.
There are several different contexts:
- what do we do in DNS clients (netcat, wget, et al)?
- what do we do in the DNS server?
 (I presume skip the record and return NXDOMAIN.)
- what do we do in the DHCP server?
 Set it to an arbitrary string? (Not nice for my aplication, since if
 multiple clients send bad strings, you end up with multiple hosts named
 'bad')
 Sanitize the name? (We were doing this poorly, hence my patch.)
 Clear the name? (I need a hostname generator anyhow, so this would
 be nicest for me.)
- what about dumpleases?
 (I'd planned on fixing the DHCP server, and relying on that. But it would
 be proper to apply the same fix...)
- what do we do in the DHCP client?
 (this is the only one that's been 'solved'.)

My question was whether generalizing the 'sanitize hostname' code made
sense. Seeing that bug, I suspect that's a better function to use for
at least everything but the the DHCP server and dumpleases.

Thanks,
Isaac Dunham