Meeting tonight?

[Jay Hannah]

Oh, looky there -- the calendar says it's been 4 weeks again.

I've got a huge work rollout tonight, so I might be able to come out, but probably won't be able to pay much attention.

Any key holders going to be there to open the door?

Anyone have anything they want to present?

Anyone want any of these books?

https://twitter.com/deafferret/status/580211563920932864

Thanks,

Jay Hannah
Project Lead / Programmer
http://www.iinteractive.com
Email: jay.h...@iinteractive.com
AOL IM: deafferret
Mobile: 1.402.598.7782
Fax: 1.402.691.9496

[David Knaack]

I'll be there to open the door by 6:30.

--
OMG!Code http://code.omahamakergroup.org
---
You received this message because you are subscribed to the Google Groups "OMG!Code" group.
To unsubscribe from this group and stop receiving emails from it, send an email to omg-code+u...@googlegroups.com.
Visit this group at http://groups.google.com/group/omg-code.
For more options, visit https://groups.google.com/d/optout.

[Jay Hannah]

On Mar 24, 2015, at 5:11 PM, Matt Payne <Pa...@MattPayne.org> wrote:
> May I please have the scala and jQuery cookbook?

Yup. See you tonight. :)

Anyone else want any of these?

[Daniel Pfile]

I'll take "Internet Core Protocols."

[James Harr]

We ended up just talking shop about a variety of topics. Here's a few links. Add to them.

http://www.infoq.com/presentations/SPDY - SPDY / HTTP2 (concepts apply to both, protocol on the wire is a little different)

https://aphyr.com/posts/281-call-me-maybe-carly-rae-jepsen-and-the-perils-of-network-partitions - Call me maybe, explorations of distributed databases when you net split them.

http://blog.ipspace.net/2014/03/ipv6-only-data-center-deployment.html - Building a green field datacenter using IPv6 only (except for an IPv4 gateway at the edge). As it turns out, if you're doing a new deployment, this is the way to go.

http://packetpushers.net/author/rwhite/ - Russ White's blog posts. One of the best sources of practical information about the internet.

[Jay Hannah]

On Mar 24, 2015, at 5:11 PM, Matt Payne <Pa...@MattPayne.org> wrote:
> May I please have the scala and jQuery cookbook?

Matt: I left those at the Maker Group for you. David Knaack might have borrowed the Scala book.

On Mar 25, 2015, at 8:29 AM, James Harr <james...@gmail.com> wrote:
> We ended up just talking shop about a variety of topics. Here's a few links.

I was hoping people would contribute links to

https://github.com/omg-code/omg-code/blob/gh-pages/index.html#L126

But I can't seem to get anyone else to update our website? :)

j

[James Harr]

Also, somewhat related. IPv6 ND (ARP equivalent) wreaks havoc on ethernet multicast forwarding. Warning: this gets into the weeds a bit.

http://inconcepts.biz/~jsw/ipv6_nd_problems_with_l2_mcast.pdf

http://blog.bimajority.org/2014/09/05/the-network-nightmare-that-ate-my-week/

This is something that's likely going to have a work-around in switching today, and eventually IPv6 ND will get reworked to not be so horrible on multicast reduction.

To unsubscribe from this group and stop receiving emails from it, send an email to omg-code+unsubscribe@googlegroups.com.

Visit this group at http://groups.google.com/group/omg-code.
For more options, visit https://groups.google.com/d/optout.

--
OMG!Code http://code.omahamakergroup.org
---
You received this message because you are subscribed to the Google Groups "OMG!Code" group.

To unsubscribe from this group and stop receiving emails from it, send an email to omg-code+unsubscribe@googlegroups.com.

[Rob Townley]

James, interesting reads.   IPv6 has proven difficult for MIT Computer Science.  

mbone video conferencing used multicast over twenty years ago, but still today  few vendors seem to get it.

To unsubscribe from this group and stop receiving emails from it, send an email to omg-code+u...@googlegroups.com.

[James Harr]

Mbone was actually multicast, not IPv6. You might be thinking of 6BONE.

Mbone worked well because it was small scale. Multicast routing at internet scale still doesn't work well if at all. The only real hope would source specific multicast, but you still run into the issue of managing a large list of receivers in core internet routers, and adding flow state to core routers is a very bad idea (scale, dos exposure). On top of that, if you join a multicast group, that join has to make its way via control plane signalling to the source. There are timers and delays involved (my mcast gets rusty here) and that can take minutes.

That, and no one knows multicast very well because it's not widely used (me included), so troubleshooting across multiple networks is near impossible in practice.

So, multicast in small administrative domains works well. Anything larger and it's easier/more cost effective to do multicast at the application layer. Then what you have is a Content Delivery Network (CDN) that does data replication. CDN administrators have a lot more visibility into problems, it scales well on top of unicast networks (which everyone knows about), and is faster and easier to work around failures. I forgot to mention that if a multicast router fails over, it can be a long time before you get your streams back flowing.

And this isn't really a vendor problem, multicast is just hard to scale and there is very little demand for it.

The IPv6 + MLD problem was a poor protocol design decision made by people that were thinking of reducing throughput in individual links without considering the switching requirements to do so. It's all in the TCAM, when a packet comes in, you have nano seconds to make a forwarding decision. Table lookups need to be fast. TCAM can do a lookup in one clock cycle, but it's expensive.