I’m taking a short trip around Japan next year, so I needed something light and small to carry with me, and Google/Samsung have just released this interesting machine at a very reasonable price:
- Very lightweight
- Seems fast so far (running ChromeOS)
- Not sure I’ll get used to the touchpad
- Confused about the difference between Android and ChromeOS, and why they didn’t just put Android on it
- Samsung invent yet another $%!? proprietary power plug
I’m using ChromeOS at the moment, but since that requires an always-on internet connection and is basically a fancy full-screen web browser, I’m going to wipe it and put either Ubuntu or Fedora on it shortly.
The interesting thing about the hardware is the Cortex-A15 processor at the heart of the machine. This supports hardware virtualization, and although the patches haven’t even landed in the upstream kernel yet, the hope is that this could run KVM at a reasonable speed.
Also, 2GB of RAM is nice.
Google are definitely “going Altavista” on us technical users. Their search results are becoming increasingly annoying and useless for technical queries. Some time back they changed searches so that broad match was the default. Now, even quoting search results to force exact match doesn’t seem to help. Google Instant could do with being taken outside and put out of its misery.
I remember Altavista going down the tubes.
Anyone wanna start a new search engine? I hear there’s money to be made …
For about a year I’ve been using Google Alerts to track people talking about software projects that I’m interested in. It’s a free service (up to a certain number of searches) and I’ve found it to be very effective.
Here are my current alerts:
So if you get a message from me a few days after you mentioned guestfish on some random phpBB out there, now you know why.
Google’s Chrome browser is trying to take an interesting approach to running native i686 and x86-64 code safely in a sandbox (without needing code signing, so unlike ActiveX).
Essentially they compile C or C++ with a modified gcc that generates a limited subset of machine code which can be easily validated at runtime. This paper and web page explains the precise details:
This code (after validation) can be loaded from a completely untrusted source into one of the web browser’s tabs (most tabs run as separate processes). The untrusted code cannot make syscalls directly or write over any trusted parts of the same process — this is enforced by the validation step. But because it is native machine code (or a large but safe subset thereof) it still runs at just about full native speed.
To perform operations (drawing graphics etc), the untrusted code can only call a limited set of trampolines into the trusted part of the process. These are called “syscalls” but aren’t the same as ordinary OS syscalls. They are more like safe procedure calls between threads in the same process. This “syscall” trampoline is described here:
This is a nice implementation of some old ideas: trusted compilers and code validation have been around for many years, for example on the Burroughs computers and in the JVM respectively.
I just got a Google Nexus S, and I can report it’s a very competent phone. Hardware is excellent all round — there’s really nothing I can complain about. It’s the first time I’ve used Android, and it’s also the first software I’ve seen which “gets” the iPhone and is as good or in a few cases a bit better than iOS.
Firstly, there is no search bar at the bottom of the page any more. You can’t refine your search after deciding that the first page of results wasn’t good.
Secondly, it’s buggy. Enter key does not reliably perform a search. You can’t reliably go back to get a previous “instant page” of results.
Thirdly, it doesn’t interact well with browsers. Safari at least always gives you the dialog about how “you’ve typed something, do you really want to close the window”.
Worst of all you don’t seem to be able to permanently disable it and go back to the old style.
I really wish Jeff Dean and Sanjay Ghemawat wouldn’t keep claiming they “invented” MapReduce.
And taking out patents on it.
Because I was taught MapReduce as a grad student in 1994 (10 years before the bollox above) by a number of people. And I’m quite sure it wasn’t a new idea, even then.