HN Gopher Feed (2017-11-26) - page 1 of 10
Some chip makers have hidden latency and jitter issues from common
141 points by based2http://www.badmodems.com/
floatboth - 5 hours ago
Apparently this is about DOCSIS modems.
based2 - 6 hours ago
userbinator - 5 hours ago
Judging by the picture on the main page, I thought it would be
about them overheating and catching fire...Until Intel released its
Puma 6 chip Broadcom dominated the modem chip market. The Broadcom
chip does not have latency or jitter issues.As much as I hate them,
Broadcom dominates the market because (as long as you don't ask for
source code or Linux or datasheets...) their modems and other
networking devices generally work. I'm guessing the cause of this
jitter is because Intel is trying to do more in software vs. the
dedicated hardware Broadcom uses.The DoS on the "Security Issues"
page is even more unusual, unless these modems are actually more
than just modems and include a router/NAT too?
revelation - 5 hours ago
In what world do WLAN & Bluetooth generally just work? Those are
pretty much the remaining bastions of horrible software support,
and for no good reason whatsoever.Even the graphics vendors have
got their shit together. That's how hopeless Broadcom is.
bsamuels - 5 hours ago
bluetooth will generally "just work" when using two devices
with the same manufacturer's chipset (this is not a rule by any
measure though).in my experience, many bluetooth hiccups are
caused by manufacturers having alternate readings of the
bluetooth spec, which is a massive, hard to swallow behemoth.if
you have two 1st party BTLE devices, you can expect to see
bluetooth at its best. for example, apple airpods + iphone are
probably as good as youll ever see a bluetooth pairing.
madeofpalk - 4 hours ago
To be fair, AirPods + iPhone are using more than standard
Bluetooth - Apple has their magic sauce on top to make things
work super smoothly especially for the initial pairing.
ztjio - 55 minutes ago
You can still use AirPods with non-Apple Bluetooth devices.
The pairing process isn't as braindead easy of course, but
they do work. That implies they adhere to the standard at
least to some degree.I personally only use my AirPods with
Apple products so I can't speak to whether they perform
less well on regular devices, but, I somehow doubt it.
Beyond pairing, it seems the only remaining secret sauce is
in the relay between the two earbuds and coordination
between the to for which one actually connects to the
source device. None of that matters for the BT spec.
jimrandomh - 5 hours ago
tl;dr: The Intel Puma 6 and Puma 7 chipsets have a performance
issue in a part of the packet processing that is specific to
TCP/IP, which went undetected for awhile because "ping" packets are
not TCP/IP. Modems that use the affected chipset will have high
jitter under load (especially load consisting of many short
connections), and will suffer denial of service when sent traffic
with many different port numbers. Additionally, under a previous
firmware (but according to a commenter on
on the most recent firmware), a timed process would briefly block
throughput every 2 seconds, adding to jitter issues. DSLReports has
created a tool which checks for this at
jlgaddis - 4 hours ago
ICMP is protocol 1, IP(v4) is protocol 4.ICMP(v4) is carried
over IP(v4), but it is not IP.
bbayles - 3 hours ago
This is a bit scrambled.Both IPv4 and IPv6 have a spot in
their headers to tell what type of traffic is being carried.
This is a 1-byte integer called a protocol number.The TCP
protocol is protocol number 6. UDP is 17. ICMP is 1. All are
carried over IP.Both IPv4 and IPv6 have another spot in their
headers to identify their version. This is where 4 and 6 go,
but it's not the same as the protocol number field.Similarly,
Ethernet has a protocol identification field, "EtherType."
IPv4 is number 2048 (0x0800); IPv6 is 34525 (0x86DD)
ZWoz - 3 hours ago
You are mixing different versioning systems. ICMP is protocol
1 inside Internet Protocol, where TCP is protocol 6 and UDP
is 17 . Internet version (as it named in rfc790) is
defined with four first bits, numbers 1-3 unassigned and 4 is
our IPv4 Technically ICMP is layer 4 protocol, but
defined as integral part of IP 
nalllar - 2 hours ago
It initially affected TCP, UDP and ICMP. After a firmware "fix",
it doesn't affect ICMP but still affects TCP and UDP.
colanderman - 50 minutes ago
That site does also have a forum topic discussing alternatives to
the SB6190: http://badmodems.com/Forum/viewtopic.php?f=9&t=5
dogruck - 5 hours ago
I?ll admit that I was initially confused by the link going to a
picture of a burning modem, instead of an article. Stupid me.
Just follow the navigation links on the left.
Animats - 4 hours ago
Yes. Site is awful. Result there are important. The site is so
bad that it makes the results look less credible.
cesarb - 4 hours ago
To me, it looks like a typical website design of the 90s: a
left navigation frame, with the contents in the right frame. In
fact, according to the
tags, that site was made with
FrontPage, as was common on that era.
Animats - 4 hours ago
It's not the technology, it's the awful graphics.
existencebox - 1 hours ago
This is a bit of a rant, and a bit of an open question of "is there
really nothing more I can do."I'm more than a little furious that
after being sold a product that explicitly fails under load, the
best we have in terms of consumer protections is a class action
that, if my history of class actions holds true, will accomplish
little to nothing to either punish intel or compensate the
consumers.For the last year I've been putting up with network
issues that I thought were a result of the house I moved into, when
now I find out that the sb6190 I spent a silly amount of money on
has been degrading an equally not-cheap internet plan for the
entire time. And now my best approach is "deal with it, spend 200$
more."Here's hoping consumers in the EU are more protected, for us
in the US I don't have my fingers crossed.
ztjio - 49 minutes ago
The SB6190 is a $95 product right now. Plus, it's just DOCSIS
3.0. You can get decent DOCSIS 3.0 modems for ~$50 if you don't
need maximum speeds. If you've got under ~300mbps connection, you
could replace it pretty cheaply. For example, a Motorola MB7220
is about $50 without any discounts. Even a 16x4 is about $70.
These are Broadcom based modems.
maxyme - 4 hours ago
If anyone doesn't know, always avoid Motorola/Arris Modems. I
remember a few years ago Arris tried to update my modem to add ipv6
support. Ended up having 2 months of constant issues where incoming
connections to ports forwarded would cause the router to drop all
outgoing packets until the incoming connection closed... 2 months
later Comcast rolled back the update.
bri3d - 3 hours ago
Motorola/Arris modems can have 3 major chipsets (TI, Broadcom,
Intel), each with a multitude of firmware and configuration
updates delivered on a per-carrier basis, so this is a _very_
sweeping generalization. I'd recommend researching a modem much
more in depth than just the manufacturer name on the front.
solotronics - 6 hours ago
would this show up under an IPERF3 test? I would think so but just
curious if someone with one of these modems can check. Also curious
if this shows up testing between switch ports on the modem or if
it's just IP outside the NAT.
colanderman - 1 minutes ago
I just used this tool recommended on the DSLReports forums, to
check DNS reliability: https://www.grc.com/dns/benchmark.htm100%
across the board with an SB6190. So? either the test missed an
issue, or my SB6190 has been fixed. (Latest posts in the forum
 seem to indicate that at least Cox is working with Intel on a
floatingatoll - 5 hours ago
The dslreports forums have been studying this for a year or so,
and should have those details available.From a practical
perspective, if your modem is on the list, your apps and browsers
will load pages slowly, with weird delays all over the place (due
to DNS being Most Impacted by the Puma problems). Replace ASAP.