The Turris Omnia is some very nice hardware, and built by people who are very much devoted to free software (nic.cz, the same guys who brought us the Bird routing suite and KnotDNS). It's a little bit overkill for a home router (it's got massive amounts of memory and a fiber interface), but if you're fine with spending over 200 euros, it's an excellent choice.
Alpine looks pretty cool. A sane init system, musl instead of libc, a decent package manager, what more could you want?
A couple of years ago, I inherited a proprietary ARM board with a large number of GBE NICs, an unmaintained vendor kernel, and the worst userspace you can imagine (and I know you can imagine a lot). I spent half a day trying to build an Alpine userspace for armhf and get it installed on the board.
I finally gave up. It took me 20 minutes to set up debootstrap under qemu, another 20 minutes to coerce debootstrap into using sysvinit instead of systemd. Tar.gz, scp, replace the root filesystem, and the board is running Debian Jessie userspace.
I haven't looked at Alpine since then. Is there now a convenient way to build a custom Alpine root filesystem?
University campuses? Conference halls? Hotels? Youth hostels?
(MU-MIMO is about increasing performance with multiple simultaneous clients, unlike plain MIMO, which only increases the throughput to a single client.)
Stop buying routers. Instead get a Raspi and USB wifi adapter capable of master mode.
The Pi has a single 100Mbit Ethernet hooked off the USB 2.0 bus. You're putting both the Ethernet and the wifi on the USB, which is going to get congested.
A typical home router has one or two gigabit Ethernet ports hooked directly to the SoC, with one of the interfaces connected to an internal manageable switch. It has one or two WiFi interfaces connected to a high-speed, low-latency bus. The WNDR3700 is a good example of the type of hardware people like to run OpenWRT on.
As far as I am aware, there is no cheap, hackable board that has the kind of connectivity you need for decent WiFi router.
This whole thing came about precisely because people running open source software on their routers were using channels that are only legal in Europe and Japan, thus causing interference with other equipment that's licensed to use that spectrum in the US.
The report originally cited by the FCC doesn't say anything about open source firmware. As far as we can tell, the interference was caused by devices running proprietary software that either was buggy or had been modified to not comply with the local regulation. If you know otherwise, please share your sources.
First, stateless configuration you just kill off with extreme prejudice.
That's up to you, of course, but you still need RAs -- DHCPv6 doesn't distribute a default route, it relies on RAs for that.
Second, broadcast is indeed essential for almost every protocol that runs over ethernet (because ARP)
but that doesn't stop us from turning off everything but the bare essentials. Almost all enterprise gear knows
how to let ARP (or ND) and DHCP through while blocking everything else.
So you kill DNS-SD over mDNS (Apple's Bonjour)? No printer discovery, no discovery of streaming media servers, no IPTV?
you can also convert all broadcast which you DO allow back out of APs to unicast on the RF side
That's actually a good idea, since multicast over WiFi is horribly inefficient.
They did away with private addressing (site-local) "because it breaks the openness of the internet and firewalls"
They did away with site-local addresses because they couldn't agree on the definition of a site (is your home network a site, or is it part of your ISP's network?). They've been replaced with ULAs, which are easier to understand, simpler to administer and simpler to program with.
The standard has changed so many times in the last 10 years
A few unused features have been removed (v4-compatible addressing, site-local addressing, partial support for MTUs below 1280), but the specification has been mostly stable for a good 15 years.
They did away with IPV4's simplistic subnetting and supernetting,
No, subnetting is still there. The only difference is that all leaf networks are supposed to have the same size, which is supposed to make administration easier.
introduced EUI-64 addressing
Yeah, that was a stupid idea. It's been replaced by RFC 7217 (stable private addresses).
first job of a wifi engineer that cares about not draining batteries is to start to turn off AP propagation of all broadcast and multicast traffic.
Unlike IPv4, Ipv6 doesn't work at all if multicast is filtered -- things like stateless autoconfiguration, DHCPv6 and even neighbour discovery run over multicast.
An IPv6 network is much easier to set up properly. Check out the HomeNet stuff, where you just chuck a bunch of routers together more or less randomly, with connectivity from cable and DSL and 4G, plus a bunch of wifi routers, and it all Just Works.
HDMI is a synchronous interface — video and audio data use up fixed parts of a frame's time. One might almost say that's it's just a digital mapping of an analog television signal.
DisplayPort is a packetised interface — video, audio, and whatever else you might want to carry over it can be sent at any time. Because of that, it's a little more expensive to implement (you need more hardware in the device), but it's immensely more flexible: you could carry multiple low-resolution video streams over a single port, or 3D video, or multiple audio streams, or even something else (IP, SMS, whatever) with no hardware changes.
There's a lot of multipath-related work being done right now, at the IETF, within OpenWRT, and independently.
We've been working on providing multiple routes automatically (disclaimer -- I'm a co-author). As to actually making use of the multiple routes, the solution that currently works best is MP-TCP, a set of kernel patches that allows TCP to use multiple routes simultaneously, with no modification to applications. Other solutions are SHIM6, which works below the transport layer, and Multipath Mosh, which works at the application layer.
I'm pretty confident we'll be able to have most of this stuff enabled by default in mainstream Linux distributions by the end of the year.
I got yelled at for recommending to my students to buy the course text online vs. going through the bookstore ($40-$60 bucks vs $250 at the bookstore for the same text).
Getting yelled at by stupid or corrupt administrators is part of the job.
College isn't about learning anymore, it's all about making money.
College is what the lecturers and the students make it. If you've got tenure, then you're pretty much untouchable, precisely so that you can take a stand without fear of the repercussions. (Of course, if you haven't got tenure yet, then tread carefully.)
Yeah, what I'm saying is we have been able to do exactly that (web configuration) with the various TomatoUSB modded firmwares for years along with a bunch of other stuff (VPN, etc).
I pushed my router to dd-wrt a while ago. At the time, I liked the UI on dd-wrt better than openWRT. I also noticed some issues on my specific hardware for OpenWRT. How do they stack up?
They're very different beasts.
DD-WRT is a single, monolithic image, similar to a vendor firmware but with more features. What is available tends to be well integrated into the GUI, but if a feature is not available, you're pretty much out of luck (unless you're willing to install software by hand).
OpenWRT, on the other hand, is a package based system: there's a base system and an extensive set of optional packages that you may install. It used to be the case that the OpenWRT GUI was not very good, but it has improved a lot in recent years, and I now find it fairly usable. Of course, not all packages are well integrated with the GUI.
I'd recommend going with OpenWRT. The base system should be reasonably easy to understand, and you'll be able to easily install extra software when you find that you have unusual needs.
I have been running native ipv6 and whatever other modern stuff on my ASUS RT-N16 via TomatoUSB for many years. So uh... What took you dorks so long?
OpenWRT has had support for native IPv6 for as long as anyone can remember. However, the support wasn't native, in the sense that it required some knowledge to configure properly.
With the current trunk (and this snapshot), you can configure things like DHCPv6 prefix delegation, DHCPv6 relaying, proxy-ND and so on over the web interface -- and it just works. (Famous last words.)
complaints on various forums that usually have "minor" bugs like "5GHz doesn't work" or "wifi randomly quits working after a day or two"
Just to prevent people from getting the wrong idea -- OpenWRT is fully functional and rock solid on a lot of 802.11n hardware (including 40MHz support). I haven't played with 802.11ac yet.
I don't recommend the WRT54G(L) for OpenWRT hacking -- it's a little short on RAM and seriously short on flash. There's a lot of much nicer routers available, with more RAM, more flash, 802.11n and gigabit Ethernet. Somebody else in this thread mentioned the WNDR3700v2, which I'm very happy with (but check the board revision -- the v1 doesn't work well). The successor is the WNDR3800, which is the same board with more RAM.
Maybe it's just because I'm unfamiliar with MathML, but this seems like a *very* verbose way of writing equations.
Yeah, it's pretty horrible. The only way to write (presentation) MathML is to generate it automatically from a sane input format, either a GUI or something like LaTeX.
We'd still deal with the inconvenience if it were universally supported by browsers -- but 15 years after MathML was conceived, it still isn't.
We've been waiting for math rendering support in HTML for slightly over 15 years (MathML came out in 1998, and there was HTML 3 math before that).
We've given up. Both the scientific and the higher education communities are using PDF almost exclusively, and our respective userbases (fellow scientists and students) have accepted PDF as the format of choice. At the same time, PDF support in browsers and on tablets has become good enough to make that a reasonable proposition.
But yeah, let's write blog postings about why MathML is not dead, it only smells funny.
This is an individual submission, not an IETF working group draft, and does not appear to either be proposed for an IETF wg draft or to be in the RFC Editor's queue. In short, it has nothing to do with the IETF.
While registering to the conference, have attendees fill in a form with the two questions "Are you a government employee, and if so in what quality" and "Are you a journalist, and if so, in what newspaper(s) do you publish?"
The people that you want to attend will be happy to have a name tag saying "Government employee, University of so and so". The people who feel the need to hide their affiliation are probably the ones you want to be escorted by security.
Slashdot Mirror
The Turris Omnia is some very nice hardware, and built by people who are very much devoted to free software (nic.cz, the same guys who brought us the Bird routing suite and KnotDNS). It's a little bit overkill for a home router (it's got massive amounts of memory and a fiber interface), but if you're fine with spending over 200 euros, it's an excellent choice.
Alpine looks pretty cool. A sane init system, musl instead of libc, a decent package manager, what more could you want?
A couple of years ago, I inherited a proprietary ARM board with a large number of GBE NICs, an unmaintained vendor kernel, and the worst userspace you can imagine (and I know you can imagine a lot). I spent half a day trying to build an Alpine userspace for armhf and get it installed on the board.
I finally gave up. It took me 20 minutes to set up debootstrap under qemu, another 20 minutes to coerce debootstrap into using sysvinit instead of systemd. Tar.gz, scp, replace the root filesystem, and the board is running Debian Jessie userspace.
I haven't looked at Alpine since then. Is there now a convenient way to build a custom Alpine root filesystem?
who needs this much headroom in Wifi?
University campuses? Conference halls? Hotels? Youth hostels?
(MU-MIMO is about increasing performance with multiple simultaneous clients, unlike plain MIMO, which only increases the throughput to a single client.)
Stop buying routers. Instead get a Raspi and USB wifi adapter capable of master mode.
The Pi has a single 100Mbit Ethernet hooked off the USB 2.0 bus. You're putting both the Ethernet and the wifi on the USB, which is going to get congested.
A typical home router has one or two gigabit Ethernet ports hooked directly to the SoC, with one of the interfaces connected to an internal manageable switch. It has one or two WiFi interfaces connected to a high-speed, low-latency bus. The WNDR3700 is a good example of the type of hardware people like to run OpenWRT on.
As far as I am aware, there is no cheap, hackable board that has the kind of connectivity you need for decent WiFi router.
This whole thing came about precisely because people running open source software on their routers were using channels that are only legal in Europe and Japan, thus causing interference with other equipment that's licensed to use that spectrum in the US.
The report originally cited by the FCC doesn't say anything about open source firmware. As far as we can tell, the interference was caused by devices running proprietary software that either was buggy or had been modified to not comply with the local regulation. If you know otherwise, please share your sources.
You need to be looking into open protocols, and implement them using free and open source software:
https://www.ietf.org/proceedin...
First, stateless configuration you just kill off with extreme prejudice.
That's up to you, of course, but you still need RAs -- DHCPv6 doesn't distribute a default route, it relies on RAs for that.
Second, broadcast is indeed essential for almost every protocol that runs over ethernet (because ARP) but that doesn't stop us from turning off everything but the bare essentials. Almost all enterprise gear knows how to let ARP (or ND) and DHCP through while blocking everything else.
So you kill DNS-SD over mDNS (Apple's Bonjour)? No printer discovery, no discovery of streaming media servers, no IPTV?
you can also convert all broadcast which you DO allow back out of APs to unicast on the RF side
That's actually a good idea, since multicast over WiFi is horribly inefficient.
They did away with private addressing (site-local) "because it breaks the openness of the internet and firewalls"
They did away with site-local addresses because they couldn't agree on the definition of a site (is your home network a site, or is it part of your ISP's network?). They've been replaced with ULAs, which are easier to understand, simpler to administer and simpler to program with.
The standard has changed so many times in the last 10 years
A few unused features have been removed (v4-compatible addressing, site-local addressing, partial support for MTUs below 1280), but the specification has been mostly stable for a good 15 years.
They did away with IPV4's simplistic subnetting and supernetting,
No, subnetting is still there. The only difference is that all leaf networks are supposed to have the same size, which is supposed to make administration easier.
introduced EUI-64 addressing
Yeah, that was a stupid idea. It's been replaced by RFC 7217 (stable private addresses).
Very Few large deployments.
Google? Facebook? Comcast?
first job of a wifi engineer that cares about not draining batteries is to start to turn off AP propagation of all broadcast and multicast traffic.
Unlike IPv4, Ipv6 doesn't work at all if multicast is filtered -- things like stateless autoconfiguration, DHCPv6 and even neighbour discovery run over multicast.
An IPv6 network is much easier to set up properly. Check out the HomeNet stuff, where you just chuck a bunch of routers together more or less randomly, with connectivity from cable and DSL and 4G, plus a bunch of wifi routers, and it all Just Works.
The IETF Homenet Working Group
Non-official Homenet HOWTO
I don't understand why people want DisplayPort
HDMI is a synchronous interface — video and audio data use up fixed parts of a frame's time. One might almost say that's it's just a digital mapping of an analog television signal.
DisplayPort is a packetised interface — video, audio, and whatever else you might want to carry over it can be sent at any time. Because of that, it's a little more expensive to implement (you need more hardware in the device), but it's immensely more flexible: you could carry multiple low-resolution video streams over a single port, or 3D video, or multiple audio streams, or even something else (IP, SMS, whatever) with no hardware changes.
I agree, except for the "they (singular)"
Both singular "they" and "he" are perfectly fine neutral pronouns in English. Thackeray writes:
while the English version of the Universal Declaration of Human Rights has:
There's still no opportunistic encryption in HTTPS. Does that mean I'm going to have to buy a TLS certificate for my printer every year?
There's a lot of multipath-related work being done right now, at the IETF, within OpenWRT, and independently.
We've been working on providing multiple routes automatically (disclaimer -- I'm a co-author). As to actually making use of the multiple routes, the solution that currently works best is MP-TCP, a set of kernel patches that allows TCP to use multiple routes simultaneously, with no modification to applications. Other solutions are SHIM6, which works below the transport layer, and Multipath Mosh, which works at the application layer.
I'm pretty confident we'll be able to have most of this stuff enabled by default in mainstream Linux distributions by the end of the year.
I got yelled at for recommending to my students to buy the course text online vs. going through the bookstore ($40-$60 bucks vs $250 at the bookstore for the same text).
Getting yelled at by stupid or corrupt administrators is part of the job.
College isn't about learning anymore, it's all about making money.
College is what the lecturers and the students make it. If you've got tenure, then you're pretty much untouchable, precisely so that you can take a stand without fear of the repercussions. (Of course, if you haven't got tenure yet, then tread carefully.)
Yeah, what I'm saying is we have been able to do exactly that (web configuration) with the various TomatoUSB modded firmwares for years along with a bunch of other stuff (VPN, etc).
Proxy-ND? Stateful DHCPv6? Somehow I doubt it.
I pushed my router to dd-wrt a while ago. At the time, I liked the UI on dd-wrt better than openWRT. I also noticed some issues on my specific hardware for OpenWRT. How do they stack up?
They're very different beasts.
DD-WRT is a single, monolithic image, similar to a vendor firmware but with more features. What is available tends to be well integrated into the GUI, but if a feature is not available, you're pretty much out of luck (unless you're willing to install software by hand).
OpenWRT, on the other hand, is a package based system: there's a base system and an extensive set of optional packages that you may install. It used to be the case that the OpenWRT GUI was not very good, but it has improved a lot in recent years, and I now find it fairly usable. Of course, not all packages are well integrated with the GUI.
I'd recommend going with OpenWRT. The base system should be reasonably easy to understand, and you'll be able to easily install extra software when you find that you have unusual needs.
I have been running native ipv6 and whatever other modern stuff on my ASUS RT-N16 via TomatoUSB for many years. So uh... What took you dorks so long?
OpenWRT has had support for native IPv6 for as long as anyone can remember. However, the support wasn't native, in the sense that it required some knowledge to configure properly.
With the current trunk (and this snapshot), you can configure things like DHCPv6 prefix delegation, DHCPv6 relaying, proxy-ND and so on over the web interface -- and it just works. (Famous last words.)
complaints on various forums that usually have "minor" bugs like "5GHz doesn't work" or "wifi randomly quits working after a day or two"
Just to prevent people from getting the wrong idea -- OpenWRT is fully functional and rock solid on a lot of 802.11n hardware (including 40MHz support). I haven't played with 802.11ac yet.
I don't recommend the WRT54G(L) for OpenWRT hacking -- it's a little short on RAM and seriously short on flash. There's a lot of much nicer routers available, with more RAM, more flash, 802.11n and gigabit Ethernet. Somebody else in this thread mentioned the WNDR3700v2, which I'm very happy with (but check the board revision -- the v1 doesn't work well). The successor is the WNDR3800, which is the same board with more RAM.
Maybe it's just because I'm unfamiliar with MathML, but this seems like a *very* verbose way of writing equations.
Yeah, it's pretty horrible. The only way to write (presentation) MathML is to generate it automatically from a sane input format, either a GUI or something like LaTeX.
We'd still deal with the inconvenience if it were universally supported by browsers -- but 15 years after MathML was conceived, it still isn't.
We've been waiting for math rendering support in HTML for slightly over 15 years (MathML came out in 1998, and there was HTML 3 math before that).
We've given up. Both the scientific and the higher education communities are using PDF almost exclusively, and our respective userbases (fellow scientists and students) have accepted PDF as the format of choice. At the same time, PDF support in browsers and on tablets has become good enough to make that a reasonable proposition.
But yeah, let's write blog postings about why MathML is not dead, it only smells funny.
This is an individual submission, not an IETF working group draft, and does not appear to either be proposed for an IETF wg draft or to be in the RFC Editor's queue. In short, it has nothing to do with the IETF.
I believe the generally accepted term is "Nobel Laureate".
Becker is actually a laureate of the Nobel Prize in Economic Sciences, which is not the Nobel Prize.
While registering to the conference, have attendees fill in a form with the two questions "Are you a government employee, and if so in what quality" and "Are you a journalist, and if so, in what newspaper(s) do you publish?" The people that you want to attend will be happy to have a name tag saying "Government employee, University of so and so". The people who feel the need to hide their affiliation are probably the ones you want to be escorted by security.