Profile cover photo
Profile photo
Dave Taht
2,461 followers -
singer, pianist, guitarist, writer, experimenter, theorist, hacker, maker, faker - in no particular order, on any given day. And I also write code.
singer, pianist, guitarist, writer, experimenter, theorist, hacker, maker, faker - in no particular order, on any given day. And I also write code.

2,461 followers
About
Dave's posts

Post has attachment
Maybe someday I'll stop getting choked up when I see this, but it hasn't happened yet.

https://www.instagram.com/p/BQtNTk4Brqp/

Post has attachment

Post has attachment
The NSA must have been happy for 10+ years running. Also the KGB, GHCQ and a zillion bad guys.

http://seclists.org/oss-sec/2017/q1/471

Hi. My name is dave. "Hi, dave".
And I've now gone a month without checking facebook once.
"Yea, dave!".
And I feel fine. If it wasn't for my faith in emacs and irc, I'd have not got to this day....

Post has attachment
Awesome new #bufferbloat fighting tool enhancement of the month: Now flent can not only measure flows out of band, but the tcp metrics themselves, directly - we can peer into the cwnd without a packet capture!

https://github.com/tohojo/flent/pull/91#discussion_r101904474

Maybe I'm just dumb. Every compiler, lib writer and their mother goes around creating "the worlds fastest memcmp" routine - which checks for less than, equal, or greater than, often does a fixup at link to call a pretty big routine to do this.... Why is there no "memeq"? Is the answer so obvious than I'm missing it?

What if all you care about is two segments of memory are identical? And you are comparing short stuff (like, in my case, fixed length 8 or 16 byte ipv6 prefixes)?

bool memeq(x,y,const size);

should I just cast something to a structure and do struct a == struct b? Does the compiler figure out that all I want is memeq?

memeq(prefix1,prefixb,16)

should boil down to "xor" on an arch with 128 bit registers. Or two xors and an or on 64 bit ones. Or four xors and 3 ors, on 32 bit ones. But I don't think it does.

Post has attachment

I have been dealing with all sorts of issues around expiring #ipv6 address
lifetimes and routes of late, mostly in lede-project, but also discovering
holes in coverage in mainline linux tools also.

Recently I did a:

ip -6 route replace blackhole fd89::1/128 proto 50 expires 600

This was a "in case I have a coronary while trying to fix this flood
upstream, expire the blackhole so the next sysadmin still alive can
deal with it" idea.

It didn't work - expires never made it to the kernel, as shown my ip
monitor nor by expires expiring anything. Which caused much puzzlement
later after my blood pressure came back to normal and my brain started
working again.

Now, off to poke at iproute. Joy.


Post has attachment
The ironic thing, is I can't find chips 2,3,6. They are in the house somewhere....

http://www.taht.net/~d/theironicthingaboutthisisthaticantfindchips236.svg

Post has attachment
Wait while more posts are being loaded