Blog‎ > ‎

H2, Gantt Chart, MegaMIMO, IAM, Outlook, Myth of RAM, Google & Netflix CDN Maps, CA trust

posted Oct 1, 2016, 12:11 AM by Sami Lehtinen   [ updated Oct 1, 2016, 12:12 AM ]
  • Probabilistic Programming & Bayesian Methods for Hackers
  • HTTP/2 Connection Coalescing - Nice post about explaining TCP / HTTP/2 basics. Sure, this will also create situations where things won't work and it'll be joy to debug. I love stuff which adds extra layers of obfuscation and code which can break down in multiple ways. I just love simple and working methods.
  • Kept wondering how people can expect software to be in production, when requirements specification isn't even ready when the launch date is. Somehow this is the root of so many projects, and make things just utterly ridiculous. But this just as it is, has been and will be in future. Then customer is dissatisfied because project is delayed, because they haven't made key decisions or haven't even ordered something which is vital for project completion. I wonder what kind of people it requires, if they simply don't get it that if you extended something on your Gantt Chart critical path a lot, it will affect everything after that point. Sure, there's the ultimate way to fix this, let's skip testing completely and put it straight to production because there's time pressure. Hah, and then they complain about problems. Which wasn't of course advised by the supplier. Surprised? Anyone?
  • MegaMIMO seems to be kind of lame solution. It's simple to block base stations using same frequency from sending at the same time. But how about making TrueMIMO where all antennas in the network would be optimally used for transmission and receiving. This should be possible, and as far as I understand some CDMA networks have been doing this for a long time. So there's only one virtual access point and all antennas are attahed to it. There's no need to roam, because the one access point is everywhere. It uses all antennas (or all usable antennas) to receive and transmit as well as allocates frequencies in it's own space totally dynamically. - That's what I would call optimal. That MegaMIMO stuff was just very simple optimization to existing multi AP configuration. Perfection would be "single virtual" access point with like hundreds of antennas around the large building complex. But all this is so obvious there are other limitations forcing other solutions.
  • Attended one day lesson about Identity & Access Management (IAM). As expected, it didn't contain anything new at all which I wouldn't have known before.
  • Worked a lot, consolidated a ton of different credentials documents into centralized credential storage. The only right way to do it.
  • Hello Microsoft, what kind of Administrators you've got? I think your staff makes me laugh. I've received message from Outlook, because Outlook is on blocklist. Lulz. 550 5.0.350 Remote server returned an error -> 550 SC-001 (BAY004-MC3F5) Unfortunately, messages from 104.47.2.236 weren't sent. Please contact your Internet service provider since part of their network is on our block list. You can also refer your provider to http://mail.live.com/mail/troubleshooting.aspx#errors. So Microsoft guys are blocking traffic from Microsoft. That's awesome. Thank you for also immediately permanently failing that message and not making it soft fail. This is just to ensure that messages will get lost. Keep up the good work guys.
  • The Myth of RAM - Very very nice post. Size does matter after all. No surprises really, but many developers don't acknowledge these well known facts. Don't forget to check parts II, III, and IV.
  • Really nice Google Infrastructure view with Maps. Nothing new I guess. But it's nice to see them publishing the data. Similar unofficial map for Netflix CDN.
  • Age long boring discussion with developers. About "official trusted certificates" versus fingerprinting... They claim that 'official certificate' is better. I say bullshit. I say that fingerprint is much better than official certificate. But who gives a bleep. Doesn't really matter, as with most of security things, nobody actually cares. If it seems to work, or is simple to do. That's the way to get it done. We also know that sometimes doing things right, is hard or very hard. It's much better to accept that things aren't being done right, and that's the generally accepted way and so what. It makes everyone's daily life much easier than following annoying and time wasting official and correct protocols or processes. Many problems are best solved, by not considering those as problems. It's just the way things are, not a problem.