smart power grids
I found this news item, which implies that modern power grids are becoming too smart because they provide accurate real-time prices for electricity, leading to problems due to load spikes when many consumers try to make use of time intervals with very low prices. While the stated problem is very realistic and important to be solved in order to facilitate more widespread use of renewable energy, I cannot agree with their conclusion that the power grid is too smart. In fact, I would say it is not smart enough to properly allow the best possible distribution of currently available energy.
The problem is that the model employed does not properly fit the reality of energy networks. It assumes that at a given time, there is a certain price for energy, with an unlimited supply being available for that price. But in reality, the advertised price is only valid for a limited amount of energy because if that amount is exceeded, a spare power source would have to be used. So the proper way to distribute the energy would be if the energy provider created sell orders on an energy market for the exact amount it can sell for the given price. Then, a fraction of the consumers who try to make use of the low price would have their buy orders fulfilled at the intended price, and the others would have to wait for another opportunity to get cheap energy.
There are energy markets for this, but at least in Germany, access is not available to consumers. But I wonder: If consumers can get smart equipment which uses real-time energy pricing data in order to find the best time to start the dishwasher, why don't they get smart equipment which brokers for a good energy deal on an energy market automatically?
emulating SO_REUSEPORT in Linux userspace
The SO_REUSEPORT socket option can be a useful feature. Not only for multicasting applications (where it supposedly came from), or for load balancing servers (which seems to be what most people are interested in regarding this flag). It also allows for clean and simple solutions for some problems which are approached in a rather hacky way otherwise. For example when updating servers, it allows to easily avoid unavailability between stopping and starting the server, as the new server can just be started before stopping the old one.
Unfortunately, it is a non-standard socket option. And for the widely spread Linux kernels, it has not yet been integrated (apparently there was a patch to do so, but it wasn't accepted). And even if it was, it would still mean to sacrifice portability to use it.
However, POSIX allows to emulate something similar with a little effort. The listening socket descriptor can be shared between multiple processes, and it can be passed to new processes through a local domain socket. Thus, while kernels cannot always be depended on for allowing to reuse ports, applications can share these themselves for reuse.
Using boost::asio as a socket abstraction layer, I built a small library to automate this, together with an echo server implementation to demonstrate its use. The API is loosely modeled after boost::asio's acceptors (async_accept works the same way, but to start listening, an asynchronous async_connect call is needed as the process has to wait for another process to share the socket). You can check it out yourself here. And be sure to let me know your comments.
multi-language T3BLOG
The mere existence of this post is some kind of an irony, as it is practically the English language counterpart of of the recent German language post which focused on how I managed to implement the blog for a multi-language site while not having to translate every post.
However, a bit of web server log analytics told me that figuring out how to use the T3BLOG extension for multi-language sites already seems to be a major incentive to visit my site, and this tells me at least two things:
- The patch to allow this might be interesting also for English language readers
- As time permits, it might make sense to further refine this adaption (for example, relating the categories to a specific language would be a nice-to-have)
Anyway, for those interested, here is the patch again. It adds a dropdown to choose a language for the posts in the backend, and filters the posts according to the current language setting at the frontend.
freelancer.com account security
Since I began using it, the online freelance job marketplace freelancer.com had always been using HTTPS to secure the users sessions. This makes sense. After all, the payments for the freelance jobs are processed through that platform (and you are required by their TOS to do it that way).
With their recent layout revamps, however, some links on the site were apparently implemented as absolute links leading to the unsecured HTTP version of the site. Thus, when accidently clicking on these links, the session cookie is sent unencryptedly, allowing man-in-the-middle attacks.
While it is modern nowadays to ignore such issues until the damage was done, the issue becomes acute for example if you try to get insurance against online theft of the funds while having to tell the insurance companies that you use unencrypted HTTP for managing payments.
Luckily, at least for Firefox users, there is a solution: HTTPS Everywhere allows you to have HTTP URLs automatically rewritten to HTTPS. Using the following ruleset, this plugin will make freelancer.com secure again:
<ruleset name="freelancer.com">
<rule from="^http://([^@:/]*)freelancer\.com/"
to="https://$1freelancer.com/"/>
</ruleset>
