DJ suspended for playing music

[Big Words ]

Classic Gold have just suspended Tony Blackburn for "violating station policy" by repeatedly playing Cliff Richard tracks. Head of Programmes Paul Baker wrote in an email to Blackburn:

"As I said on Monday, we might carry out research on him, but for now we have a policy decision that he doesn't match our brand values, he's not on the playlist, and you must stop playing him."

Although BB is not a huge fan of either Blackburn or Richard, we can understand their point here. Speaking for ourselves, we never let a disc even pass over the waiting tray of our hifi without first running it by the focus group and checking that it aligns with our brand values ("Is this a load of hoary old prog rock rubbish? Great, put it on!"). "Musical appreciation" was just some boring concept mentioned in a school lesson that we ignored once. Cliff is a well-known Christian and a virgin, whereas Paul Baker and Classic Gold are clearly corporate suckers of Satan's scaly cock (cf. Bill Hicks). It's like having bacon butties at a barmitzvah.

"We don't play any Shakin' Stevens either," said Managing director John Baish in a blatent attempt to claw back listeners.

Update, 2004-06-24: Blackburn is back and Cliff will now be played "as much as the Beatles, and we play the Beatles quite often" (...adds John Baish helpfully). It's a stunt, and a pretty cupid one - how else did the full text of the station's internal emails get into the original report? You might think this would be cause for us to retract our statements above, but we feel it further underlines that Classic Gold's managers remain Beelzebub's fellatio handmaidens. Unfortunately, in this case we also got a taste of that wormy jism. Remember: for an endless diet of bland MOR to nullify the empty void of your daily existence, tune in to Classic Gold (or, if you don't have digital, Radio 2).

Oi, where's the ****in' bar, John?!

[Big Job ]

Another long time Linux peeve resolved: for the past six months or more, my rxvt terminals have lacked a proper mouse cursor. The standard cursor shape was, and should be, the I-bar. Sometime around one of the many RPM upgrades, it suddenly became a white arrow with, annoyingly, a selection zone under the arrow itself rather than at the apex, thus making it very difficult to accurately select text. File under "Minor irritations that will slowly drive you nuts".

I couldn't trace the problem, and I couldn't put together a sensible search string to produce a lead via Google either (not surprisingly, "X11" and "mouse pointer" throw up a lot of references). Rxvt didn't seem to contain any changes to the cursors. Xterm had the same arrow, but in black on black, which was even less helpful (marginally). Judging by the lack of complaints about this problem, it may only manifest itself in this way on my system.

Finally, I struck lucky. It seems that the XFree86 team introduced themable mouse cursors in 4.3.0 (because ghod knows, we don't have enough themable programs compared to all those practical, fully-featured Linux apps). So excited were they by this innovation, they didn't even wait to finish off the actual themes before releasing it, with the result that the default set of icons contains several bugs and omissions. (I can't understand why everyone embraced the X.org code fork, given XFree86's careful stewardship and maintenance.) Fortunately, even though the XF86 guys seem a little vague on the subject themselves, you can revert to the original behaviour. Add the following to your .Xresources file:

Xcursor.core: true

Then either update your current resources (xrdb -merge .Xresources) and restart your X clients, or simply restart your X session. Presto, the bar is back.

Of course, if I'd read the Xfree86 4.3.0 release notes, I might have seen this little tidbit (and the actual mention was small). But I'm not in the habit of scouring pages of docs every time I run an APT update. I guess sudden subtle changes in expected behaviour are one of the delights we have to relish in the brave new world of Fedora and constant improvement.

Other bubbles

• Those exciting Xcursor resources in full, should you wish to work with the system.

WebSphere 5.1 plugin issues

[Big Job ]

The Linux cheerleaders are hugging themselves with glee ever since IBM came down so heavily on their side. Personally, the more I use IBM software, the more convinced I become that their support is either a) an underhand Microsoft-funded sabotage attempt (paranoid conspiracy theory par excellence of every Linux hacker) or b) a curse. Of course they embraced Linux - did you ever try using AIX??

Here's a fix for the NSAPI (Sun ONE/IPlanet) HTTP plugin in WebSphere Application Server 5.1 that you won't find on the IBM support site (heck, you're lucky to find the IBM support site).

We're using WebSphere with Zeus Web Server on Solaris/SPARC, which also supports NSAPI extensions. Using the latest 5.1 plugin (libns41_http.so), we got the following error in the Zeus errors file after starting the instance:

Run-time exception error; current exception: GSKDBException
        No handler for exception.

I think this is a generic catch-all message (IBM's only advice on the subject refers to Apache and can only helpfully suggest "restarting the web server"). Using "truss -fl" on the main Zeus parent process prior to starting the instance, I found the following in the output immediately before the error was logged:

26380/7:        open("/opt/WebSphere/AppServer/etc/plugin-keyrdb", O_RDONLY) Err#2 ENOENT
26380/7:        close(-1)                                       Err#9 EBADF
26380/7:        time()                                          = 1087484502
26380/7:        write(2, " R u n - t i m e   e x c".., 59)      = 59

The file it should be opening is plugin-key.rdb (for want of a period, lossage ensued). The same thing will also happen with the plugin-key.crl file. (I use truss a lot, but it puts me in mind of the saying, "when your only tool is a hammer, everything looks like a nail". Trouble is, a hammer is often the only effective remedy for much of this shit.)

To fix this, simply create symlinks with the incorrect names in the WebSphere etc/ directory, pointing to the correct files:

# ln -s plugin-key.rdb plugin-keyrdb
# ln -s plugin-key.crl plugin-keycrl

(And yes, just for you lovely people, I'm prepared to brave the rigours of IBM Support - "We need full debug output and your complete life story before we can start examining your PMR" - to report this issue.)

Update, 2004-06-25: With a little help from IBM, following a brief sightseeing tour of several blind alleys, I have found a fix. The fix is in PQ86671 for IHS 1.3.12 on Solaris (Possible security exposure with IBM HTTP Server regarding a denial-of-service attack at the SSL record/protocol layer), which contains GSKit 4.0.3.345 (later than the versions in either WebSphere 4.0.7 or 5.1.0.4). Confusion stems from the fact that the IPlanet plugin on Solaris uses an older version of GSKit than the other plugins, which ships with an older release of IHS; make sure you download the right version of the IHS fix.

Who's your daddy?

[Big Words ]

Seen ahead in this morning's commuter jam: a large Jag with personalised number plate also featuring the Playboy bunny symbol, and on the rear windscreen a car sticker from The Prisoner gift shop at Portmeirion:

who is number 1?

It's not entirely clear that the driver appreciated its full context.

Printing Holy Grail discovered

[Big Job | Big Picture ]

There are currently lots of fine things being said about the HP Photosmart 7xxx range, particularly their amazing black & white printing capability, but mainly by those lucky people who have already taken a chance on one. The rest of the world, or at least the part that skulks in Internet photographic fora, seems content to continue swapping tips on unclogging Epson heads while pointedly ignoring the HP users raving about their great printers. I'm delighted... overjoyed... ecstatic with my 7660 (ex-Epson users tend to be like that when they finally switch to any other brand of frickin' printer) under Linux, so here are a few hints on making it work well for photo printing.

The USB side of things should work fine, and if you're using a recent distro or an up to date foomatic database, there should be a driver entry under the CUPS setup.

In the GIMP print dialogue, select PostScript Level 2 as the printer model then locate the PPD file. Assuming you've already configured CUPS, this should be in the /etc/cups/ppds directory (under Red Hat/Fedora). Change the print command options; delete the -oraw part and add -o'PrintoutMode=Photo' to ensure the printer goes into photo mode (you can get a list of other options with lpoptions -l). Most of the other dropdown menus will be grayed out, mainly because the printer autodetects the media type and tailors the ink output appropriately. If you're using the HP no. 59 grey cartridge then either colour or greyscale will work for B/W images. Toned images seem to come out well with the #59 cart in colour mode (the standard #57 cart supplies the coloration).

I had several months of excellent prints from my P7660, but then I began to see some really poor results. Banding, consisting of broken ink layering, started to appear across the images. At first, this only seemed to occur with Ilford Classic Gloss paper. Then the problem spread to other gloss papers. Finally, even prints on matte paper were affected. I tried running cleaning cycles. I tried new cartridges. I tried other brands of paper. Finally, in desperation, I ... I ... booted Windows. Installed the bundled HP driver (um, after three hours and a hundred reboots to update IE to the right level) and ... got a perfect print (despite the rather limited image-handling software that HP supply). Humph.

Back to Linux for some deeper investigation, now I knew the printer worked correctly. I noticed that Ghostscript was preprocessing the output at 300dpi, whereas the printer is capable of 1200dpi. So I went into the CUPS configuration and set up a new photo-specific queue for the same device, with the output explicitly set to "1200dpi, Photo" rather than "Controlled by Printout Mode". Using this queue, Ghostscript processed the output at 1200dpi resolution and the result was a perfect print. Once more.

I have no idea why the output quality changed and degraded so subtly, but I suspect it may have something to do with the behaviour of either CUPS or Ghostscript being altered by one of the many Fedora updates. Theoretically, the default settings should have worked fine (and indeed did for a while), but at least this way I know the output will always be handled at the maximum resolution no matter what.

Some good things about the P7660:

• B/W printing, as already mentioned, is superb. I have no idea how it compares to traditional wet prints (some claim it equals or betters them, depending on your skill level in the darkroom), but they're Good Enough For Me.
• Colour rendition appears almost exactly like the onscreen image without any tweaking or profiling, although doubtless proper calibration would make up the remaining few percent difference. However, the 7660 tends to print lighter so you usually have to reduce the gamma by 10-20% and maybe boost saturation slightly.
• There are almost no head clogging problems at all. Even if it happens, you can simply replace the cartridge to get a new head (which partly explains their phenomenally high pricing, although "gouging the consumer" would seem to be a major component too).
  Contrast this with the Epson 830 (or indeed any Epson), which clogs within the first seven days or five seconds of a dust molecule landing on the casing, whichever comes sooner, and then stays clogged despite copious cleaning, floods of Windolene, servicing, sanding or a damn good bashing with a heavy mallet prior to tossing the useless heap of junk in the bin (pant, pant, pant...).
• Even printing direct from a memory card produces acceptable snapshots.

Minor issues: continually swapping cartridges for different types of printing (there are varying optimum combinations for text, colour photos and B/W photos) is a pain, as is the printer's insistence on running a calibration test (with paper) each time a new cartridge of a particular type is inserted.

Some people feel that the more expensive 7960 is the one to go for, because its imaging algorithm is superior and its ability to hold three carts at once means a greater greyscale range and less swapping. That may be true, but if you can't afford the high end model, don't sweat it; I doubt you'll be dissatisfied with the budget range.

Other bubbles

• LinuxPrinting.org: font of all knowledge.

The new law of supply and demand

[Big Words ]

Forgive us for banging on about not being able to buy anything we like in the shops, but The heavy price of seeing food as a commodity by Andrew Anthony in yesterday's Guardian contained an apposite observation:

"Large corporations like British supermarket chains do not define demand in terms of what people want, so much as what their customers can get elsewhere. And as supermarket competition is driven not by quality but price, what they can get elsewhere is the same, only perhaps cheaper."

This goes not just for supermarkets but practically every other major retailer. Certain "leading" high street photographic stores come to mind. As do those defaced Aber pubs we bemoaned the other day. Thanks, Adam Smith and all the proles.

Go home then

[Big Noise ]

All About Eve's grassroots Top 40 comeback campaign has hit the buffers, as their new single, "Let Me Go Home" stalls at no. 52. Given that a band only needs to sell about fifty copies to reach no. 1 these days, this is a pretty bad showing. Presumably, if BB hadn't bought a copy they might be at no. 62 instead. They had it all so well organised too, apart from one key element:

"Right, so we've got the fan mailing list set up?"
"Check!"
"The CDs are pressed, they're on the way to HMV?"
"Check!"
"Web site ready, downloadable sample?"
"Yep!"
"And we've got a decent song..."
"Err...I thought you were writing that?"

Unfortunately, "Let Me Go Home" has the staying power of water on a duck's back. It's a three minute slice of goth nu-rock that suggests AAE thought they could ride the Evanescence breakthrough. But the chorus is indistinguishable from the verses and none of it can be recalled five seconds after the player has moved on to the next track. My Glamorous Research Assistant also noted that it disappointingly lacked any qualities of their previous work; we're not asking for another song about dancing gypsies, but at least give us something equal to the great and much-maligned "Ultraviolet". Which begs the question...why do we need this?

Nostalgia for pub bores

[Big Deal ]

Just added, the Nostalgic guide to Aber pubs, a lament for those drinking dens of yore and how they've been steadily eroded or ruined ("modernised"). "Had to leave to get back to Swansea in time, so couldn't go to see whether the pubs are the same too," said Telsa of her recent Aber trip. Probably for the best, chuck.