Off Topic/Meta: How is this poorly tagged? The C tag covers C++ too
I suspect that it’s due to the missing book tag, which I’ve added.
Thenks! When adding tags, C and C++ belong to the same tag (with, I believe, C#). IMHO C and C++ should have different tags.
According to the tag descriptions, C# belongs in dotnet, but not in c.
You’re right. It’s Objective-C the one included in the same tag as C and C++.
You downvoted the post for being poorly tagged when the tag did not exist? Oh god.
I wonder why the MacBook Pro 15" didn’t get updated with the new trackpad; only the 13" did.
I read on arstechnica that the 15" refresh isn’t coming until mid 2015 when the quad core processors finish.
I’m happy to see that the MacBook Air isn’t being left behind (yet).
Maybe / hopefully a bigger update is coming there sometime this year …
I’ve slowly been working my way around to believing that hosts is an anachronism. Everything should be DNS. If you want to override DNS, then actually override DNS. I have too many machines to be editing files on. :) even if you only have one machine, running a DNS cache on local host is the bomb.
What do you recommend for doing that these days? The last time I messed with DNS was with PowerDNS.
Still using bind, but fooling about with unbound more.
Unbound is pretty darn good. I don’t use it configured with dnssec though, so I can’t speak to that.
I’ve been running dnsmasq locally for a few months, seems to work well. Not had any issues with it.
Agreed. The only thing I end up in /etc/hosts these days is my hostname matching an entry for lo0, though even that may not be necessary anymore. (Remember back in the late 90s when X would hang for multiple minutes starting if it couldn’t resolve itself? Weird.)
Does that work with DNSSEC?
I really kind of wish Linux had the flexibility in DNS that OS X offers (guess you don’t hear that phrase a lot): It lets you register resolvers specific to subdomains, which is allows stuff like pow to work.
HN has this feature and I use it a lot when I remember a specific article from last week but I can’t remember it’s title. Incase anyone’s wondering the URL is:
There’s other providers; but if your datacenter does not provide ipv6 yet you will need to just tunnel it in.
Most give you a /64 to start with; that’s a lot of ip’s to work with. If you’re talking just a vps this should be doable; the scripts on tunnelbroker.net should help you set everything up.
For reliability, I’ve had significantly better results with SixXS than with HE.net’s tunnel services. Test thoroughly either way, of course.
journeysquid, I think that depends on your endpoint entirely. I have had the opposite experience with SixXS.
I was a SixXS user starting in 2006. I previously used HE.net but I changed due to their fire-walling policies that have since changed.
Let me start by saying SixXS uses credits (or did). So you sign up and you request a tunnel and after 2-3 weeks of having your tunnel up you have saved up enough credits to ask for another tunnel; or a /48. I was naive I did not understand the stupidity of this; virtual currency basically but you cannot buy any credits.
So having pops go down from time to time; you get docked credits; it costs you credits to change your pop (and you get a new ip allocated sometimes). I found this behavior to be harmful to the consumer (me) and pointless so I went back to he.net where they gave me everything i wanted without making me wait 3 weeks; or save up credits or nag me when my isp is down for a day.
I do hope that SixXS has become better since then; but looking back it’s like some kind of virtual currency sham to penalize (me) for wanting ipv6.
SixXS is the reason I’m asking my ISP for a native address; my tunnel goes down very frequently. I’d say once a week or so. That might not seem like a lot, but it causes quite a bit of disruption. Enough to move my home server off my network and to disable IPv6 on most of my devices. Once the ball was rolling it wasn’t long till I was disabling AICCU on my router. It’s very disruptive to try SSHing into a server and have it fail because I’m using an IPv6 address and my tunnel is down. Again.
Maybe you had better results but SixXS disappointed me quite a bit. Also, once I saw connectivity problems, I’d sign in and see the tunnel in green; it seems the system rarely detected any problems but I could see multi hour long outages on the graphs; all would fall to zero.
I’d like to try HE at some point but I’ve been using a tunnel for around a year now. My ISP should be implementing native IPv6.
I hate to bring this in from HN, but this is pretty important.
paulirish (supposedly of the chrome team) writes:
This is a new UI experiment that’s deployed to a small fraction of users. We’re looking at a few key metrics to see if this change is a net positive for Chrome users. (I imagine it may help defend against phishing).
He goes on to present his opinion against it.
The comment about phishing is interesting, although I was under the impression Chrome already does this to some extent; notice the domain is written in black text and the rest of the URL is a faded gray which emphasizes the domain. Users may not be aware of this though.
Two things I really like about Go:
1) The go tool chain, which includes go fmt, go get and godoc. Built in tools like that make managing go code painless. They are so well integrated into my workflow, removing them would be detrimental to my productivity.
2) It’s “just works” approach. For instance, static binaries mean I can send someone a binary and they can just run it. If someone sends me source, it’ll just compile and just run. That’s a refreshing change from more dynamic languages, particularly Python, which I’ve often found require some work to get them to run.
These two alone are so key, if Rust doesn’t have them, it’s fancy pointers won’t be enough to get me off Go.
While Go has first mover advantage, Rust gets to learn from other languages.
Rust’s version of go get will be implemented as cargo by Yehuda Katz and Carl Lerche who previously built Ruby’s Bundler.
Likewise there is rustdoc and rustfmt. All three are currently slated to be launched with Rust 1.0, but there’s the possibility of rustfmt missing 1.0.
Since Rust is ideally replacing C / C++, the language designers are trying to make the runtime as small as possible (e.g. garbage collection and heap pointers are moved into a library) and defaults to static linking. There’s no reason to believe that a Rust binary will depend on shared libraries unless explicitly compiled that way.