News for hackers. Stuff that matters.
Re:TCP Anycast – Don’t Believe the FUD [pdf] by justizin on Thursday May 28, 2015 @02:31PM (#9620120)
Super informative, love the detail in the slides!

Would be more impressed if the IPv6 plan wasn't:

  The plan consists of being dead by the time customers demand v6.
Why aren't these brave heroes of anycast leading the charge to ensure it works with a protocol suite that is now something like a decade old?

I don't expect the story to be:

  We are just as confident with this as what we've been running for years on IPV4.
But, maybe:

  Except for the bits that we rely on custom hardware for, it's looking really promising.
Getting IPv4 allocation from ARIN, even if it is conserved by anycast, isn't going to be possible forever. I've been meaning to get ahold of an ASN and a small block for just this reason.

Therefore, the techniques outlined here may not be usable to new companies or on new networks in a couple of years. Even if I can get IPv4 from my hosting provider, I need ARIN/RIPE/etc.. for anycast, afaik.

Still, great read! :)

Re:TCP Anycast – Don’t Believe the FUD [pdf] by namecast on Thursday May 28, 2015 @02:38PM (#9620173)
It's a protocol layer thing, mostly down to "what prefix sizes can I reasonably expect other providers to accept?"

For IPv4 the answer was a /24; for IPv6 the answer seems to be a /64 at this point, and burning an entire /64 to get a couple of DNS servers (the typical use case for anycast) to resolve slightly faster is overkill. It's also probably a great way to annoy your RIR when they ask 'how the hell did you burn through your initial IPv6 delegation with only two hosts' ;)

EDIT: I meant /48, not /64! Whoops-a-doodle.

Re:TCP Anycast – Don’t Believe the FUD [pdf] by detaro on Thursday May 28, 2015 @02:51PM (#9620268)
It's also probably a great way to annoy your RIR when they ask 'how the hell did you burn through your initial IPv6 delegation with only two hosts'

Since /56 seems to become the "easily available standard size" for allocations, using a /64 doesn't seem like to much of an issue, especially since larger orgs should easily get /48s? (But are /64 really accepted as routes? I would have imagined that to be also limited to /56 or /60 or so)

Re:TCP Anycast – Don’t Believe the FUD [pdf] by namecast on Thursday May 28, 2015 @03:05PM (#9620369)
Depends on who's doing the accepting ;) I actually meant a /48, mea culpa!

I'll leave this here:

https://labs.ripe.net/Members/dbayer/visibility-of-prefix-le...

It's from 2010-ish but still sadly relevant.

Re:TCP Anycast – Don’t Believe the FUD [pdf] by detaro on Thursday May 28, 2015 @03:29PM (#9620548)
Ok, yes, "wasting" a /48 is a problem ;)