Bandwidth charging makes it to the US!

I hear that the big American ISPs are about to start charging for bandwidth. Now, of course, a lot of the rest of the world has been doing this for years, but at least some of the US ISPs seem to be going a bit mad, with quite expensive plans allowing 5GB per month, and other such silliness. I am reminded of the Asimov story involving a horribly over-complicated and expensive tax system; when this was pointed out to the necessary-plot-element-dictatorial-leader-with-no-access-to-economists, he switched over to poll tax.

On the one hand, really low bandwidth restrictions are over-the-top, and may actually effect the way people use the Internet out of fear of going over (the over-use charges are silly, too); I doubt Google is overly impressed.

On the other hand, though, no-cap Internet service probably isn't really a great idea. Inevitably, a small number of people will use the lion's share; UK ISPs give figures along the lines of 5% of users using 95% of capacity. It really isn't terribly fair that your Internet access slows to a crawl because your neighbour is downloading a few terabytes of porn, and bandwidth is not free, even for the ISPs.

It'll be interesting to see how this one plays out. Hopefully, it won't scare people off the Internet too much.

Embarrassing iPhone article from CNN, and Flash musings

The iPhone is on everybody's mind right now, or at least on the minds of all us obsessive Apple followers. It seems fairly certain that a new version, probably 3G, will be released tomorrow.

Here's CNN's take (on the speculative patent-based third-generation iPhone, which they think is the fourth-generation one because the second-generation one due out tomorrow has '3G' in it, you see...). Besides the talk of noted always-just-round-the-corner vapourware Wi-MAX, there's this gem:

Just how will Apple meet expectations? Using the patent application as a guide, Apple appears to be making room on the iPhone for flash memory, which means an end to Apple's standoff with Adobe (ADBE) that's kept iPhones from easily viewing a plethora of Internet videos.

Apple has said that Adobe's flash media player, which is on hundreds of other phones, doesn't perform up to Apple's standards for the iPhone.

Erm. Right. Having more Flash memory makes Macromedia/Adobe Flash faster, you see. The fourth-generation (fifth, or possibly nineteenth in CNN land) will have Java memory to make Sun's Java faster, and possibly Office 2009 memory to make Exchange faster.

Honestly, I realise that they're not a technology writer, but could they at least have someone proofread this crap?

Also, what is going on with that Flash thing, I wonder? The iPhone isn't dramatically slow, and most non-video Flash apps aren't that demanding. If I had to guess, I would say that it was more of a manifestation of Adobe's slowness to port to new platforms (no x86-64 version, yet, remember), and possibly a bit of payback for Apple's cruelly grabbing the Carbon rug from beneath their feet; the next version of Photoshop for MacOS will only be 32bit, allegedly, because Carbon (Apple's old, creaky, SDK, used by Photoshop, MS Office, MCL, various others, and for that matter, I believe, Flash) never made it into the 64bit world; they'll have to switch to Cocoa for that.

Could that be the issue with Flash-on-iPhone? iPhones don't do Carbon at all, I don't think; it's entirely possible that Adobe just didn't have a suitable plugin in time.

The trouble with video blogging

Much has been made of the potential of videos on the Internet. Thousands of people do 'video blogs', and any new web framework or whatever worth its salt will have roughly 50 narrated screencasts of how to make a blog or clone Reddit or whatever.

The thing is, does anyone actually watch these? The equivalent in text is nearly always easier to follow, quicker to read, and less bloody irritating. I've never sat through a whole episode of a 'video log', or, indeed, one of the web framework videos. And yet I read lots of normal blogs, and articles on web frameworks and similar. I seriously doubt that I am alone in this.

They're more trouble for the maker as well; you need equipment, and somewhere to put the things, and I suspect they take longer to make than the equivalent would to write. Then there's the whole 'putting video of yourself talking on Internet' psychological aspect. I know I wouldn't really be able to deal with this; I can barely look at still photos of myself without vague irritation, and cringe if I ever hear my voice recorded.

Robert Scoble, noted ex-Microsoft person who blogs enthusiastically about Web 2.0 stuff, is a prolific video blogger, producing hour-long videos where he talks. Why, exactly, this is better than an audio recording of him talking, or, indeed, than the text version thereof, is unclear, but he, and many others seem to be really, really keen on the idea. He even, apparently, films the damn things in HD; quite why anyone would want the image of Scoble going on on their screen to be a higher-resolution version is unclear. Showing off would be my suspicion.

So, am I just a weird luddite? Or is it the case that no-one actually watches these things?

Wikia Search, and a cat

Wikia, Jimbo Wales' company, launched a search engine a while back. No-one really noticed, because it was dreadful (especially the ranking; it's pretty nonsensical), but it got a bit of publicity today, for reasons I'm not clear on.

Anyway, look what happens to the header if you search for 'cat':

It's not a bad cat, I suppose, but felines of far higher grade can be found with little effort on Google Images.

I'm not sure why anyone thought that endless tiled pictures of cats (or whatever; things turn up for quite a few searches) was a particularly sensible idea. It looks pretty dreadful, generally.

BBC being even more BBC than usual

Note that four out of the five most popular items are totally absurd. That's high, even for BBC.

Reddit's wonderful comment markup

Today, I posted the following as a comment on Reddit:

Note the italics. You see, I tried to say 'mod_lisp'; Reddit evidently thought that the underscore meant that I wanted emphasis. Unfortunate on a site where weird underscored names are likely to be used often, really.

Still, I quite like it; it looks a bit like one of those funny bibles where random words are in italics, doesn't it? Try reading it aloud. MOD-proxy. Makes you sound like a mad preacher.

Oh, also; 285 milliseconds ago. Really? 285 milliseconds since what, exactly? It being submitted? It being put in the database? Who can say, really? And why the high precision? Most people limit themselves to 'seconds ago' at absolute worst...

Benchmarking the new Hunchentoot Common Lisp Web Server

Version 0.15.7 of Hunchentoot, the popular http server for Common Lisp, was recently released, with the promise of significantly increased speed, due to improvements in Flexi-Streams.

Hunchentoot has always had speed issues with actually transferring data, due in large part to the slowness of the flexi-streams library it uses, so I was curious to see what the difference would be. In previous versions, time to send a page has been more or less proportional to the size of that page.

I conducted a rather naive, simple benchmark, comparing 0.15.3 (the version I previously had installed) and 0.15.7 (the new, fast version) on SBCL 1.0.15 for MacOS 10.5, using a 2.4GHz Core 2 Duo Macbook Pro. I was largely interested in the speed of sending data, so I prepared three pieces of data to send; a 14 byte string, a 300KB string, and a 600KB UTF-8 string (something copied off Google.jp, appended with itself a lot). These were stored in memory, and I used ab to see how fast I could retrieve them. Here are the results. (Available, and possibly more readable, on a separate page here.)

Update: There are now also results for the new refactored version of Hunchentoot.



As you can see, I also ran a comparison against Apache's file serving, and PHP. Unfortunately, the long string for PHP was rather shorter, as PHP doesn't seem to like really long string literals. I ran a Hunchentoot benchmark with the same string for comparison.

The results are interesting. First, for the tiny page, you can see that the new Hunchentoot is over twice as fast. Quite impressive. But for the big document, the new version is nearly ten times faster. I was shocked by this; I hadn't been expecting such a gain. The UTF-8 document is slightly faster again; I included it to check if flexi-streams was using a special fast mode for Latin-1, but clearly it isn't.

The results for the development refactored version are also interesting. Note that the long file results are considerably slower than the better Hunchentoot result, while the UTF-8 results are the same, and the short results are considerably better. I'm not sure what's going on here, but the refactored version is still in development, so may improve.

Apache, as expected, is much faster again; nearly another ten times. PHP, interestingly, is about the same speed as Hunchentoot, as far as serving files goes. This is pretty cool; it makes Hunchentoot extremely competitive for generating dynamic webpages, as Lisp, all things being equal, should be far faster at putting pages together, processing data, and so on; the single OS process nature of a Hunchentoot server also allows trivial caching and connection pooling and so forth.

There were a few other things I'd have liked to try for this benchmark, but didn't manage. First, I wanted to try the new, development version of Hunchentoot, which has been seriously refactored and should in theory be a bit faster again. Unfortunately, common-lisp.net is currently having technical trouble, and I wasn't able to get it from SVN. Update: I got it from here. Also, I would have liked to try it with SBCL on a Linux machine; the MacOS version of SBCL still has issues, especially with threading, and I do wonder whether that slowed things down a bit. That's also the reason I only used concurrency of 10 for AB, by the way; higher can cause SBCL for MacOS to drop connections, but there is no such issue with SBCL for Linux, and it manages to maintain decent speed with hundreds of concurrent connections. I don't currently have access to a suitable Linux machine which isn't doing anything else, though.

In any case, I think that these results show that Hunchentoot is very much ready to be used for real, high-traffic sites.

Today in stupid technology news

America may be getting free wireless broadband all over the place! Complete with injected ads and content filtering (billed as 'family friendly'). I'm not sure exactly what the message is, here; no porn for poor people, perhaps? Given the high quality of most content filtering systems out there (they tend to be good at blocking out revoltingly lavicious sites like Wikipedia and BBC News), I doubt it'll be all that great.

It brings up an interesting issue, though; if the readiest source of th'Internet is rendered 'family friendly' by whatever obscure metric the provider chooses to use, with FCC approval, isn't there vast potential for abuse of the censorship? CDA 2.0, anyone?

Also, the EU is planning on shoving people gently towards IPv6 by 2010. Now, we all know that that is not going to happen. IPv6 is a convenient mechanism for breaking certain types of communication on MacOS; Erlang, for instance, tends to get confused about the loopback interface. That is all. What do we want? IPv6! When do we want it? When every single operating system, router and third party software package is upgraded or retrofitted to work properly with it, and not a moment before!

I mean, imagine. Suddenly Google vanishes for users of, say, BT, because Google is saying it does IPv6 and BT's routers are saying they can do IPv6, but Windows 7 or whatever merely believes it can do IPv6.

And finally, here, allegedly, is Windows 7. Could they take up any more space with giant start menus and toolbars if they tried? Notice, also, that the evil psuedo-menus merged hideously into the title bar from IE7 look set to become standard. When CDE (ancient, hideous UNIX window management system, beloved for many years by Sun) looks like a more attractive UI experience than Microsoft's latest, maybe, just maybe, that should give them pause?

Are you mad?

Then you'll love Intel's absurdly-named SkullTrail platform!

Basically, it's a dual-socket Xeon system for the lunatic gamer market. With 10 USB ports. Ten!

I'd be impressed if a computer based on this came in at less than 10,000, and it's clearly aimed at mad gamers, rather than anyone else. Some people have far more money than sense...

ASA versus the expensive HiFi cable industry!

I've mentioned the whole 'super-expensive magic power cable will make your audio equipment sound better/computer faster/fridge colder (really)/whatever' thing before. I am, of course, sceptical. Vendors of these devices generally claim that while results can be easily heard, they can not be objectively measured with instruments, which should be very suspicious in itself.

Anyway, the UK Advertising Standards Authority has upheld a complaint against one such company, on more or less the basis I've mentioned; audio distortion can be measured, and the company did not provide credible evidence that people even thought that the thing sounded better; proper double-blind tests would indeed be somewhat convincing, even if they would possibly undermine electronics as we know it.

Funnily enough, the offending device is one of the cheaper ones; it only costs 30 pounds. You can pay thousands for these things. It will be interesting to see what impact this decision has, though; advertising magical cable properties without substantiation is very common.

Fun iPhone app

Here's a Wikipedia app for the iPhone/iPod Touch. No, it's not a specialised browser; it actually has the whole of Wikipedia (from 2007) as a file. Not much use if you have an iPhone with a data plan, but very nice on an iPod Touch. It does, I'm afraid, take up 2GB, but such is life.

Highly recommended if you have a jailbroken iPod Touch.

Wordpress.com becomes EVEN CLASSIER - child-porn-implying ads

I've babbled at length about Wordpress.com before. Well, they've really outdone themselves now.

First, wank.wordpress.com reports that anti-Scientology blogs are showing Flash ads for Scientology. Now, quite frankly, I'd prefer not to see those anyway; I don't think advertising an evil cult which includes mad ol' Tom Cruise is in the best taste.

Oh, but you ain't seen nothing yet. It appears that they're now showing ads on 'mature' blogs; generally Google doesn't allow this so presumably they have made an exception for Wordpress.com. But what ads. Here are some ads on a perfectly normal ad about the world's largest penis (in China, apparently):

Please note the top ad. The one for 'Puberty Pictures'. Ewh, Google. Ewh. I don't think I shall be starting an adult blog on Wordpress.com...

Odd Google Talk feature - say hi!

As you can see, I now have a 'say hi' widget on my blog sidebar. It's a new Google Talk feature which will let users of your site talk to you in a web browser window. It seems quite clever, but so far no-one seems to have used it. Well, one person did, but they disconnected before I answered.

Give it a go. :)

Yahoo(!) Fail(!)

A bit of advice for all the giant web companies out there; don't allow this to happen to your major, popular, high-profile services:
Been like that for the past 20 minutes or so. I mean, do they want to lose their userbase?

Banking in the FUTURE!

I just looked at my online banking, to see this withdrawal:


Please note that this withdrawal happened next Monday; it can anticipate my actions!

Actually, of course, the withdrawal was yesterday evening, on my way home from work. I realise that banks don't really believe in weekends, but it's still a bizarre way to present it.

Have you bought YOUR chip company yet?

Everyone's doing it! Within a few days of each other, both Sun and Apple bought a chip designer. Apple's makes low-power PowerPCs, Sun's makes (imaginary, as yet) low-power x86s.

Funnily enough, before the Apple/Intel deal, it was widely expected that they would use processors from said PPC designer in their laptops. Interesting... Apple, of course, have kept their options for fleeing to another platform quite open; just about all Apple applications are still compiled for PPC and Intel, which forces developers to consider endian issues and so forth, and a move back to PPC would be particularly easy. However, Apple seems to have most-favoured-vendor status with Intel at the moment, so a move at the moment doesn't look likely. Still, Apple has been known to dump chip makers rather publicly and violently (Motorola over the G4, IBM over the G5), so maybe they're keeping their options open.

Sun's move is more obvious. They bought a low power x86 company. They have a low-power many-cored UltraSparc (T1/2/2+/3) themselves, and would no doubt like to improve it with IP from the acquisition. They might also be interested in making many-cored, low-power x86s, stealing a march on Intel and AMD.

Interesting, nonetheless. I wonder who will be next to buy a weird chip designer? I'm sure Sun is eyeing Azul...

More marvels of captioning from Wikipedia

From the world's premier collection of chairs (yes, I know I'm stealing this from the Register, but Valleywag has been doing so too, lately, so I feel justified), comes this wonderful caption:

Well, yes, so it is. Quite frankly, they shouldn't have bothered.

AMD in embarrassing admission

AMD's announced that it's bringing out a 12 core chip.

The 2010-destined six-core "Sao Paolo" processor and 12-core "Magny-Cours" will incorporate DDR3 memory and an additional HyperTransport 3.0 link. Magny-Cours will use two six-core die in a multi-chip package.

What's funny about this? Well, AMD has been harping on for years about how dreadful it is that Intel's 2 and then 4 core chips were not 'true' 2 and 4 core chips, but rather two 1 or 2 core dies in a multi-chip package. It was basically their entire marketing campaign for the current Barcelona chips. Ah, well, the shoe's on the other foot now, eh?

Intel should have an 8 core chip (a 'true' 8 core chip, as AMD would have said last Thursday) in the same timeline; maybe they can shove two in a package and call it a 16 core.

Desirable Doodads

Wants:
It's an FPGA development board. I used to love this stuff in college; it'd be quite nice to have my own one to play with.

It's the time to buy one, of course. What with the anaemic dollar, it works out at less than 100 dollars. Hmm...

Disturbing UI message of the day

You know how Apple is famed for their wonderful UI design? Well, they don't always hit the nail on the head, it would seem. From the iPhone SDK installer:
Yargh! In fact, it really means that it is installing MacOS 10.5 development tools, but it is one scary way to put it.

In fairness, though, the new version of XCode is quite the prettiest IDE I have ever used. You should see the code completion! And as for the electric brackets... stunning, simply stunning. Brings a tear to the eye.

Sadly, I do most of my proper work in Eclipse (CUSP and ErlIDE), which isn't so pretty.