Posts by Mark A 10 publicly visible posts • joined 14 Aug 2008
What worries me is that while better safeguards _could_ be built into the algorithms, they aren't because it would take another couple of milliseconds.
The other mystery is why the exchanges (or governments or regulartory bodies) don't put a stop to it by building in a 1-second random shuffle buffer to their transaction processing. Or is it that there needs to be (another) global catastophe before any action is taken...
It seems like a problem that speed (and FPGAs) means the algorithms have to be simplified, potentially removing the possibility of any reasonable safeguards being built in. If everything was throttled to a slower speed, that headroom could be used for smarter rather than faster algorithms with better stability.
What puzzles me is why don't regulators enforce this? Seems like it could achieved with a bit of buffering into the exchange with a random shuffle (based off an entropy source). Or is everyone making too much money leasing the fastest slots/lines?
Restarting every major daemon on the system sounds like a long-winded equivalent to a reboot to me.
You could argue that restarting a process is quicker, but not by much (the OS reboot should be what, 60 seconds at most), and conversely is several patches affect the same process it could need multiple restarts rather than just one.
Overall in terms of availability it seems like a wafer-thin distinction.
The article on why Linux apparently doesn't need to be rebooted to apply patches is interesting... since Unix effectively reference counts files, you can replace something and the old one just hangs around while one or more processes have it open.
The presents two obvious problems:
1. Different processes can be running different versions of a device driver for the same physical hardware - this seems like a recipe for interesting stability problems.
2. More seriously, you haven't applied the patches for real at all. Any running processes (drivers, server processes, daemons) are still running the old unpatched code, until you do the reboot anyway.
It's hard to see how this can be considered to be better than the Windows model, except to lull sysadmins into a false sense of security...
Last time I checked, signing up for tethering isn't just switching on that option on the phone, it also gets you the same data plan as a USB dongle, ie 3GB/month instead of 1GB.
More importantly if you exceed the 3GB you get charged the excess at a similar rate (~£12 for each addition GB).
However on the iPhone, if you go over 1GB, you go into a £2.35/MB charge band so that extra GB will cost you over £2K.
Can anyone explain why TVs still suffer from overscan issues? Plug a DVI into a monitor and you get a perfect 1:1 alignement. On a TV (and I've tried this with many, including my brand new Sony Bravia) and invariably the display is misaligned. Indeed NVidia's control panel has a feature to let you stretch/squeeze the secondary display visually to align it on your TV.
Also console games and set top boxes still have to leave "safe area" borders that you can't use for UI in case it's offscreen for some people. And of course this also means you're not getting 1:1 output pixels on your nice 1080p supply chain.
So what is it that TVs are doing differently (or more cheaply) than monitors in this respect? Just seems wrong that a digitial display with a digital input can't get this right...
You're confusing A-GPS with network-based location, which is based on cell-tower triangulation and very inaccurate like you say.
A-GPS is regular GPS - the only difference is how the initial GPS lock is obtained, which the trickiest part. Plain-ol' GPS has to work out which satellites are in range, and download a bunch of data (time corrections etc.) modulated over the weak GPS signal. As any N95 early adopter knows, this can take a long time and tends to go wrong when you're moving.
A-GPS contacts an assistance server over your phone's data connection and gets the info that way resulting in a near-instant lock. You've got to be careful of data charges though, especially when abroad, or if you're on some sort of "more than zero in and you pay for 24 hours" tariff.
http://en.wikipedia.org/wiki/AGPS
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
