ip.7 update on treatment of reserved addresses?

ip.7 update on treatment of reserved addresses?

[Seth David Schoen]

Hi,

I've submitted a patch a couple of times that adds more detailed
information in ip.7 about how the Linux kernel treats reserved IPv4
addresses.  (There is some discussion already there, but it's not
entirely comprehensive and up-to-date with respect to the current
behavior.)  I'm paying attention to this because I've been actively
involved in efforts to get the kernel to treat these addresses more
permissively than historical standards suggest -- which it indeed
does.

I haven't gotten any feedback on my patch on the occasions when I've
submitted it, so I thought I'd ask directly whether anyone is
interested in reviewing it or discussing it, or whether there would
be any interest in expanding the documentation on this point under any
circumstances.

Thanks!


(specifically, Linux now permits you to assign the lowest, or "network",
address on a segment as a unicast address; it has for several years
permitted you to assign addresses from within 0/8; it has for many
years permitted you to assign addresses from within 240/4; and all of
these are also accepted as valid unicast destinations -- none of which
users would assume just from reading RFCs!)

Re: ip.7 update on treatment of reserved addresses?

[Alejandro Colomar]

[-- Attachment #1.1: Type: text/plain, Size: 2207 bytes --]

Hi Seth,

On 4/13/23 01:06, Seth David Schoen wrote:
> Hi,
> 
> I've submitted a patch a couple of times that adds more detailed
> information in ip.7 about how the Linux kernel treats reserved IPv4
> addresses.  (There is some discussion already there, but it's not
> entirely comprehensive and up-to-date with respect to the current
> behavior.)  I'm paying attention to this because I've been actively
> involved in efforts to get the kernel to treat these addresses more
> permissively than historical standards suggest -- which it indeed
> does.
> 
> I haven't gotten any feedback on my patch on the occasions when I've
> submitted it, so I thought I'd ask directly whether anyone is
> interested in reviewing it or discussing it, or whether there would
> be any interest in expanding the documentation on this point under any
> circumstances.

Sorry for that.  During the last couple of years, maintenance of the
project hasn't been easy, as I'm alone doing this.  It may be that I
miss some patches, or that having seen them don't find the time to
address them.  Please feel free to resend any such patches, and ping
as much as necessary.  Only if I explicitly reject a patch would mean
I'm not interested in it, which will rarely happen.

Also, please mind that I'm alone maintaining this, and may not have
enough background to even understand the details of some new APIs, so
it is encouraged to CC any other kernel maintainers or developers
that may help reviewing your patches.  Please have a look at the
CONTRIBUTING file:
<https://git.kernel.org/pub/scm/docs/man-pages/man-pages.git/tree/CONTRIBUTING>

Thanks,
Alex

> 
> Thanks!
> 
> 
> (specifically, Linux now permits you to assign the lowest, or "network",
> address on a segment as a unicast address; it has for several years
> permitted you to assign addresses from within 0/8; it has for many
> years permitted you to assign addresses from within 240/4; and all of
> these are also accepted as valid unicast destinations -- none of which
> users would assume just from reading RFCs!)

-- 
<http://www.alejandro-colomar.es/>
GPG key fingerprint: A9348594CE31283A826FBDD8D57633D441E25BB5

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 833 bytes --]