Blog‎ > ‎

PEP 492, KaPA, XBRL, Loon, Tableau, Pwnie Express, Cloud Prices, MongoDB, Great Cannon, Containers

posted May 2, 2015, 9:00 PM by Sami Lehtinen   [ updated May 2, 2015, 9:00 PM ]
  • Checked out: PEP 492 - Coroutines with async and await syntax for Python 3.5
  • It's just wonderful how simple MS guys have made freeing up disk space on server. How hard you can make managing disk space? At least Microsoft seems to be trying hard to make it as complex and annoying as possible. I would say it's user experience, but in a really bad way. Btw. 2012 version keeps complaining that the binaries are 16 bit which is clearly a lie when binaries are from wow64 or amd64 directories. But no can do! Of course it will eventually work out, but this is again example of situation where thing that should take less than minute ends up being annoying struggle. Just as remark, I kow you can install full I don't want ton install the whole Desktop Experience package which includes tons of stuff I don't need.
  • I'm still thinking that India dropping out of is bad thing. Why? Because it's better to have even limited Internet access than no access at all. Access to Wikipedia alone can be literally a life saver! It's easy to forget that there are 4 billion people without any kind of Internet access. SpaceX, OneWeb, Facebook, Google, Qualcomm, latency, Intelsat, Project Loon, Ascenta, graveyard orbit.
  • Actually this is great question, what is net neutrality and is it discrimination or freedom? It sounds like with some topics actually freedom is used as basis for discrimination. Isn't forcing quotas also discrimination even if some people see it as gender neutrality. Afaik true gender neutrality means that gender doesn't matter, if there's quota being assiged by some statistical distribution and it's enforced to achieve "neutrality", isn't that actually discrimination? Hmm, great questions. Why there's food aid? Why some people are forced to pay for their food when others get it for free? Is that discrimation? No, I'm not actually trying to make any kind of stand here. I'm just asking questions and wondering "what's right and why" and if there's even absolute truth to such questions.
  • Checked out Kansallinen Palveluarkkitehtuuri (KaPA), jossa osana on kansallinen palveluväylä ja sen tiedonvälityskerros. Same in English: National Service Architecture which includes service and data transport layers.
  • It's also great that project Loon is progressing.
  • Reminded my self about XBRL and XBRL @ Wikipedia
  • Tableau 9.0 got again huge performance increases. Version 8.0 was already just blazingly fast, and 9.0 futher drastically improves speed of data visualization. It's just wonderful to notice how fast things can be done via proper analysis and optimization. If you compare how quickly you can pull group by results from average SQL server versus Tableau the difference is just huge.
  • Made extensive investing cost comparison analysis, including all trading fees, work time, taxation and other factors. If it's better to invest using direct stock purchases, and using which stock broker, low cost index funds or ETFs. Currently I'm utilizing all three options, so I also had track record of actual costs I could check out as well as compare potential differences where possible. Only thing guaranteed to happen when investing are costs, potential profits and related taxes may come or many not come.
  • OpenBazaar 0.4 released with it's new rUDP transport. Yet it seems that after start the program consumes more and more cpu and memory when time passes, some cleanup systems aren't clearly working correctly. After enough time has passed, the process silently crashes.
  • Pwnie Express - It detects active attacks, but doesn't detect passive attacks. So it won't help to many of tracking threats.
  • There are stunning price differences between computing and storage service providers like Amazon AWS, Microsoft Azure, Google Compute Engine, UpCloud, Hetzner and OVH.
    Here's AWS vs UpCloud comparison by UpCloud. As said there are huge differences on "cloud service provider" pricing. UpCloud is clearly a lot cheaper than AWS, Azure or GCE. But there are even cheaper options if you're just looking for those. No wonder OVH is Europe's largest service provider. It seems that some of the service provider market is really illiquid and some service providers charge ridiculous fees as well as provide exceptionally bad service even while charging a lot. That's the field where many of the traditional Telcos are. You'll end up paying 10-20x the market price for the service. No wonder they got these 'sales guys' talking lot of trash and then charging a lot. In many cases you won't even get full console or management card access to servers as well as you don't get proper management console for the systems. I've been talking with several of these providers and it's hard to believe how much effort they put in sales, which should be placed on automation and cost efficient processes.
  • Time for To-Do list task cleanup. Let's see how many I've completed, I've just forgotten to check completed. And how many I can now discard as 'expired' and time has just gone past those. And how many relevant tasks there will be left. This is going to be interesting. Yet it's usually very nice to notice that tons of tasks have been completed without just updating the To-Do list.
  • It's interesting to see how often nothing actually new pops up. Google Project Fi. Is just so old stuff. It's just like fully automated traffic (not only cars) or personal flight. So old stuff that you won't believe it as well as immediately obvious for everyone. Only problem is actually making it practical and cheap enough. So idea is nothing, execution is everything. All parts of this concept have been widely used, but yet have failed to gain wide spread long term use.
  • Pretty nice post comparing MongoDB and Azure DocumentDB. Blob storage part was quite compact. Great questions about blob storage would be if it's efficient and how it handles transactions. Is it also fully transactional with the main object data. If blobs are external data, and stored outside the main document, then those are clearly separate data structure and it get's even worse if those aren't handled transaction wise. In many cases where I know there's large blob data, and I don't need transactionality I immediately choose not storing large blobs in database at all. I prefer storing file id or hash instead of the data. If large blobs are stored with rest of the data every update to the document most probably leads to situation where the blob data is also rewritten to disk making updates really expensive and slow.
  • Security now 504 and TrueCrypt security audit, China Great Firewall and Great Cannon.
  • Just as I've said earlier, filtering DDoS is extremely hard. Because in case of the Great Cannon, they did just what I've said. Run the attack on high level and from distributed sources and make completely valid requests. Then it's really hard or impossible to know if this is attack or if it's just flood of actual users. Attacker can even simulate completely valid traffic. Yes, it's harder than just flooding packets, but it's also much more effective. Also everything they say is new attack blah blah, is decades old. As long as IP and TCP. There's nothing new there. Intercepting and modifying traffic has been going for a long time. As I said, I did at one network monitoring company back in early 1996. Why? Well, just because I could and I had good pre-existing toosl. So that makes me absolutely sure that many, many, researchers and tool developers had though about that stuff several years earlier. Even Trumpet Winsock allowed to dump packets, as soon as you can easily see what's there, it becomes immediately clear, that it's possible to modify the content too. I remember that some early VoiP apps allowed to masquarede their traffic into ICMP pings and into multiple other message types when required. Anything over anything, isn't anything special, because it's just bits. You can wrap or convert those into other forms, like, light, radiowaves and so on. As well as you can wrap data into existing formats just like text tv in TV signal or stereo in FM transmissions. It's obvious.
  • Yet another programmer encountered 'strange issues' with floating point arithmetic. Well, there's nothing strange about it, it's just how it is. And it has been standardized. See: IIIE 754 / ISO/IEC/IIIE 60559:2001
  • Read article about DARPA's Memex search technology
  • When every you're talking with some project, it's good to get immediate picture in mind. This is the user interface they're needing, these are the integrations, messages and message formats they will be needing, and thse APIs should be used. AS well as what kind of hardware they require for the project as well as database tables and schema. Know it all, is a great approach!
  • I want to do and accomplish things. It's always important to find a team who loves their job so much they would be doing it even without getting paid. So if it makes good money on top, that's all just a great bonus!
  • Checked out yet another e-receipt product. It was nice and technically working. Yet, what's the point? Who's getting the benefit? Customer? Reseller (Receipt Issuer)? Bookkeeping office possible receiving the receipt as part of traveling expenses? Technically it's trivial to do, as soon as there's demand. And there won't be demand before there are users and there won't be users as long you don't do anything with the e-receipt.This is the traditional new technology problem. I remember USB it was a joke for several years. Also everyone got their own implementation.
  • Reported a few OpenBazaar issues: [ 1244, 1245 ]. It's good to notice that commit 58463c2 fixed the issues.
  • Checked out new CDN provider CDNsun. They got really dense network. But I guess Akamai got plenty more POPs than they do.
  • Cinia's project C-lion got now it's own pages. Connecting Finland and Germany with direct high capacity fiber. Without passing via Sweden and Denmark.
  • OVH doesn't seem to automatically assign IPv6 address for Windows 2012 R2 servers. I'm wondering if this can be a security issue. When they let you know about the IPv6 address you'll have to enter it manually. When you enter it, Windows defaults to /64 network. But OVH is actually using /56 networks. After quick check it seems that there might be possibliy a security issue and MITM possibility here?
    Traceroute with incorrect /64 prefix:
    1    <1 ms    <1 ms    <1 ms [2001:db8:2152:7001:d3m0::2265]
    2     1 ms     1 ms    <1 ms  2001:41d0:52:7ff:ffff:ffff:ffff:ff7e
    With correct /56 prefix the first hop is different:
    1     1 ms    <1 ms    <1 ms  2001:41d0:52:7ff:ffff:ffff:ffff:ff7e
    Is this a security problem or not? I'm not 100% sure, but at least it very potentially could be. It depends what's the role of the vps3 in the traceroute is. I guess I have to ask them. I checked that they're using manual configuration for IPv4 and IPv6 on Linux boxes and there aren't any router advertisement (RA) packets being delivered on the network which could work as good hint where to forward packets.
  • Watched Atari Game Over documentary.
  • Containerization can bring new blind spots to IT security. Just fire and forget containers can cause serious security problems, just like obsolete network hardware. Excellent blog post about this topic. How this differs from most of installations? People just setup something semi randomly and when it seems to work, that's it. Then they forget it, until it stops working and they have to do something to fix it. No news there. Unfortunately.