[colug-432] IoT devices, IPv6 -- Re: ping on z/VM

[R P Herrold]

As noted after the CARPE [2] meeting (those not present missed 
a good one) last night: One takeaway was that Arduino's are 
too small in memory space for comfortable coding [heap, stack 
and stored configuration], but that the Espressif ESP8266 and 
the later ESP32 were much better for those not wanting a 
full-blown operating system (as a Raspberry Pi, Beagleboard, 
and so forth permit).  There are reasons, mostly power 
consumption, no SD card woes, and speed of blowing in a new 
set of code for incremental testing, to prefer NOT lugging a 
OS around

IPv6 came up.  The processing, and memory space requirements 
which IPv6 carries with it are probably properly out of scope 
for mainframes and big enterprise corporate (which was the 
writing prompt for the attached post), but certainly 
additionally apply to IoT and ESP8266 devices ;)  This is not 
really news -- When 'too many' ASNs were hit -- something over 
20k as I recall, big iron router ram needed expansions as well 

How say you, Rick T ?

-- Russ herrold

---------- Forwarded message ----------
Date: Tue, 15 Aug 2017 11:43:40
From: R P Herrold <herrold at owlriver.com>
To: IBMVM at LISTSERV.UARK.EDU
Subject: Re: ping on z/VM

On Tue, 15 Aug 2017, Hamilton, Robert wrote:

> Hear, hear.  +1, even.

Earlier, David Boyes: 

> > To be blunt, it’s about time a major vendor made IPv6 the 
> > easy and default case.
 ...
> > For the problem at hand, I like the idea of ping / ping4. 
> > Easy compromise and easy to implement, and makes it clear 
> > that IPv4 is deprecated and should be the outlier, not the 
> > rule.

I guess I am a 'live and let live' pragmatic Luddite.  People 
with 'lesser' networking needs will still need to configure 
office and residential networks.  The RFC 1918 blocks are 
pervasive, as is the hardware in support of such.  I do not 
know that IPv4 is deprecated much of anywhere other than as 
transit networks. Why not just prefer the mDNS / Avahi / 
Rendezvous / Bonjour / ZeroConf ... Lord save us with the 
namespace duplication?


I've had an open ticket with my 'datacenter' folks, as they 
(inadvertently and unknowingly) dropped HE out of the 'blend' 
of upstream transit, and with it, knocked out IPV6 access at 
the colo

I opened a ticket, and seemingly I was the only customer who 
has complained so far, and getting a resolution has lingered 
longer than a month.  Hurricane stills offers their 
'https://www.tunnelbroker.net/' gateways but SixXS has stopped 
offering new tunnels.  Locally ATT offers it in some parts of 
town, but not others; ditto TW / Charter / Spectrum, and if 
one has low blood pressure, call and try to get a commitment 
to provide it when ordering a new link

The answer is: we still need deterministic (and affordable) 
deployment by people of 'lesser' interest in the finer 
points of IP space exhaustion


The FOSS 'getaddrinfo(3)' implementation is well designed to: 
	allow[]programs to eliminate  IPv4-versus-IPv6 dependencies

but there was an overflow / unchecked return problem in the 
Linux implementation a couple years back.  What still lurks in 
non FOSS code?  


Reading the 'man' page, I see reference to:
	/etc/gai.conf

not present on my local RHEL 7 sources rebuild (although it 
is called out in the packaging of 'glibc'), and I suspect 
not present in ClefOS 7 either

	[root at centos-7 ~]# rpm -ql glibc | grep gai.conf
	/etc/gai.conf
	[root at centos-7 ~]# ls -al /etc/gai.conf
	ls: cannot access /etc/gai.conf: No such file or directory
	[root at centos-7 ~]# 

Getting documentation written and complete always lags 


Getting 'reverse delegation' of PTR records for DNS generally 
is nothing but a cat-fight against clueless tech's at the far 
end of the conversation.  IPv6 delegations _should_ be no less 
difficult, but ...

[I speak as a 'pervasive' and long IPv6 user and advocate [1]; 
I run ipv6 and ipv4 stacks at my office, and 'outside' there 
on an ATT Uverse link [three miles away another Uverse link I 
pay for lacks it and when calling to get it working, I get 
'dull ignorance' and no progress].  But truly, 'inside' usage 
is at best a 'geek toy' to force me to habitual usage, as a 
matter of training, rather than need].  It turns out that 
Spectrum has intermittently offered ipv6 at my residential 
gateway, so nothing I can rely on

I have no problem with 'transitioning' people to a parallel 
'ping4' [I use: ip addr show rather than: ifconfig, but is it 
really worth the extra typing?] ... but with getaddrinfo, 
perhaps the tools should be doing that, no?  It seems that 
until a technical solution is in place and not working, the 
'big hammer' of fiat re-naming is not appropriate

-- Russ herrold

1. http://orcorc.blogspot.com/search/label/ipv6
2. http://www.colug.net/carpe/