Is there a word for this?

There has to be, but I don't know what it is, and I wish to be informed. Especially you, science-y types.

So, today apparently someone came out with a study which says that CEOs with big heads are great. No, really.

Now, this sort of thing happens all the time. It's often reported as 'having a wide head makes you a great CEO!', and then boards of directors throw out all their selection criteria and just hire the applicant with the biggest head, and then they wonder why it doesn't turn out terribly well.

What's really happening in this kind of case, of course, is that the science of choosing a good CEO (or whatever the context happens to be) has become so well-developed, and so shared by everyone actually involved in choosing CEOs, that either:

a) factors which are very unimportant compared to other factors, but have a genuine effect, have their apparent effect magnified because the selection processes in use filter out all the more significant factors

so, imagine intelligence accounts for 80% of good-CEO-ness, social skills 15%, business experience 4%, and miscellaneous 1%, of which 'having a big head' is probably 0.5%. But boards all know about intelligence, social skills, and business experience, so they pick intelligent, socially-skilled and experienced people to be CEOs; the factors that account for 1% of good-CEO-ness are the only ones that now vary in the sample set, so their apparent influence becomes greatly magnified, and shallow thinkers believe that having a big head makes you a great CEO.

or, b) all of the above except that having a big head is not a determining factor in being a good CEO at all and it's sheer randomness at work: out of all the completely unimportant attributes of CEOs that vary, one which happened by complete chance to coincide with 'good-CEO-ness' in the sample set is 'having a big head'.

I'm sure this happens all the time, and I'm sure statisticians are aware of it and try to account for it and it's completely muffed by the general press, as is usually the case with Hard Science Stuff: but I don't know if there's a general term for the effect. Anyone?

Do not adjust your sets, Part 2

So, this makes diagnosis fun:

So...either there's some bug in my Fedora setup or some skiddie decided it'd be really useful to humanity in general to DoS my website. I've stuck the recommended config workaround into my Apache conf, including the logging stuff, so I guess we'll find out...

edit: while I'm still following my logs trying to figure this one out, to whoever's reading my 2007 posts on a Sony-Ericsson P900i: here's a nickel, kid...

Do not adjust your sets

Sorry if you were having trouble accessing the site for a few hours; last night my internet connection went down, and this morning I've been migrating the server from Mandriva to Fedora. MDV 2010.1 is going EOL, I don't particularly trust MDV releases from 2011 on now that everyone left and went to Mageia. I thought about going to Mageia but it just seemed to make more sense, if I was going to do a messy migration anyway, to go with Fedora - that way I can stop maintaining packages for two distros...

Everything should be more or less working again now, but please ping me if you notice any problems, I haven't run a web server on Fedora before. Login issues, missing anything, etc. Thanks!

Half-baked idea: WIMM

Well, this is one of those ideas that seems awesome in the shower, so I figured I'd write it down before I forgot about it.

If you read this blog regularly - well, see a psychiatrist, but you'll also be aware of my various frustrations in finding a personal groupware setup which meets my own requirements (which are, more or less, #0: actually works, #1: syncs with all my devices' native clients, #2: sane web interface, #3: Not Google, #4: self-hosted). It occurs to me that this might represent what those up-and-coming Young Turk capitalists refer to as a 'market niche'.

So, the idea is called WIMM (Where Is My Mind). Yes, that's a Pixies reference. No, I haven't checked for trademark issues. It even comes with a great hashtag - #wimming!

It would be a web service which gives you a very dumb and simple, personal, todo list, calendar and contact list. No whizzy complex LDAP stuff or email support, at least in v1.0 - that can come later. There would be a nice easy web interface at www.wimm.com (oh dear, I just checked and it's taken). There would be a whizzy, standards-compliant mobile version of the site which worked nice on phones and tablets. It would be able to sync with everything - F/OSS and otherwise desktop clients and mobiles OSes. This would be a really important development area.

So far, so meh...as someone is no doubt yelling at home right now (thanks for playing along), you can do that with Google already. Well, nearly. It doesn't sync great with quite everything, but it's close. However, it's, well, Google, and the server side is proprietary.

So WIMM's twist would be a) not Google and b) the server side code and API would be as public as possible. It would be F/OSS - AGPL, I guess. Rather like Firefox Sync, you would not be just allowed but downright encouraged to self-host if that's what you wanted. Self-hosting would be actively supported and very very easy for anyone who is actually capable of running their own web server.

So where's the money? Well, there's lots of people who can't self-host, and they could pay for the public server: many current operators in this area use a free/premium model, WIMM could go with that, or simply charge everyone a small flat fee to use the public server (I'm rather of the personal opinion that it's nice to have a straightforward service you pay a reasonable fee for, rather than getting it 'free' and worrying about how your free lunch is in fact being paid for). WIMM could offer support for people / organizations who wanted help with self-hosting, especially more complex setups for a small business or whatever. Obviously supporting small work groups in a useful but simple way - shared task lists, contact lists and calendars - would be a clear post-1.0 development avenue.

There's some really neat blue-sky ideas for v5.0 or whatever - how about public contacts? You could publish a subset of your contact data and let people add you to their WIMM contact list simply by entering your name, if you wanted. Or there could be cleverer ways to limit it: WIMM could generate one-time keys which you could pass out to let people add you. They could even be encoded into QR codes you could put on your business cards...yeah, crazy shower ideas, I know.

Really, though, it seems like there's got to be some space for a simple, personal, todo/contacts/calendar service that is F/OSS, actually works, isn't massively over-engineered, and syncs really well. Nothing I can find out there right now quite fits the bill: things like Remember The Milk are neat, but they're single-purpose and they're not F/OSS.

So do I start hiring developers and drawing up the plans for my mega-yacht or go back to the day job? Right, thought so - back to the day job it is...

Dear Lazyweb

I'm back on my 'personal egroupware' kick. Here's a simplified version: can someone point me at a F/OSS to do list web app? I want one that I run on my own server, not some company which runs one for me. So, basically, Remember The Milk, but the actual server code, not just a service I can use. It seems like an absurdly simple bit of code, way simpler than a Wordpress or whatever, but I just can't find one anywhere...

edit: hey, lookit that, I found one. Haven't tried it yet, but it looks like what I need.

just one 'daily log' post, then

just one 'daily log' post, to satisfy the meme and provide a flavour - I'd never be able to keep this up every day:

  • chair the QA meeting
  • news+email catch up
  • check in with jlaska and benl at red hat regarding qa team responsibilities
  • mail sriram ramkrishna some stats on F15 GNOME 3 test days
  • mail ajax to check on the status of his 'making X drivers non-critpath' project
  • mail bskeggs regarding out-of-date-ness of nouveau
  • give tao wu a heads-up that f16 beta rats compose may become tc1
  • quick mail to michael meeks regarding oracle vs. google
  • create f16 common bugs page
  • mail test@ list to ask for contributions to f16 common bugs page
  • spam devel@ list with my awesome opinions
  • mail abrt team to ask about abrt apparently not catching test crashes
  • file gnome bug on changes to tree view padding in 3.1.5
  • file gnome bug on problem with pausing/resuming in rhythmbox
  • build latest rhythmbox for f16; involves bumping libdmapsharing to latest upstream, so do that too
  • file gnome bug on rhythmbox master crashing when scanning library
  • test proposed workaround for kernel bug - 2.6.39 and later kernels cause my laptop (and others) to DoS dd-wrt routers
  • file bug on big drop in performance using kernel 3.1-rc2 on my desktop
  • discuss above bug with jwb and bskeggs, confirm 3.1-rc3 somewhat improves matters, update bug
  • further testing on f16-alpha-kde-crashing-X-in-VM bug, provide X log
  • re-schedule X test week to clear space for i18n/l10n test week to drop back, and so it won't happen while ajax is on vacation
  • respond to eclipse test day proposal
  • check in with tflink and mkrizek
  • dial in to fedora program meeting (rh meeting to monitor any fedora issues of relevance to rhel)
  • do f16 fedora-easy-karma run

still planning to add some content to the f16 common bugs page, as alpha drops tomorrow.

no, not all my days are this busy. =)

LinuxCon 2011 Day #1

I'm here at LinuxCon 2011, in Stephen Gallagher's talk on SSSD and FreeIPA, which I'm attending because I know jack squat about either of 'em. Con is running smoothly so far, the wifi works (must be fifteen T3s in parallel humming away at the back of the hotel), and the keynotes were interesting - Jim Zemlin's was pretty funny, and Jim Whitehurst did a great job of crowbarring in as many subtle jabs at Oracle as possible!

I'm heading back home at noon as I'm having a couch delivered this afternoon, but I'll be back at the con once that's done and I'll be at the gala tonight. I'll be back for Thursday and Friday...probably spending Thursday writing the presentation for Friday..

Apple vs. Samsung: a detail everyone seems to have missed

So, I got a bit interested in this Apple vs. Samsung battle this morning, and I've caught something I think everyone else has missed.

What's happened is a German judge has granted Apple an injunction preventing the Galaxy Tab 10.1 being distributed in Europe, due to alleged infringement of Apple's 'Registered Community Design' 000181607-0001. RCDs are effectively design patents: they protect the design (not function) of something, and are supposed to protect only novel and unique designs.

The RCD in question has widely been cited as 'iPad-related' or 'iPad 2-related' in the media. This does not, however, seem to be the case. There's a simple detail arguing against it: the filing date. 000181607-0001 was filed 2004-05-24, and cites a previous U.S. registration of some kind (I'm having trouble finding it) dated 2004-03-17. Supporting this is the simple fact that the RCD in question doesn't look at all like an iPad - it doesn't have the circular button below the screen, or any ports on the side at all save a charging port.

The RCD database is actually searchable; the updates on Thom's OS News story clued me in to this. You can go to http://oami.europa.eu/RCDOnline/RequestManager , enter 'Apple Inc.' as the owner, and browse through Apple's RCDs. And boy, are there a lot of them. Some of them are pretty absurd, as Thom noted.

So I went and searched through the RCD database, in descending date order, and lo and behold, I found the actual iPad RCDs. These were filed in July 2010, and - in contrast to the 2004 RCD - consist of what are clearly technical drawings and actual photos of the original model iPad. The group of RCDs all numbered 001222905-xxxx appear to cover the design of the iPad itself, and those numbered 001227003-xxxx seem to cover the UI.

Rather interestingly, the 001222905-0002 RCD - which seems to be the most basic 'this-is-an-iPad' RCD - is listed with status "Invalidity procedure pending".

There are no RCDs for the iPad 2 visible in the database. I suspect this may be because they're still in the review phase, or something. I find it extremely unlikely that Apple wouldn't actually have applied for some.

So there seems to be something odd going on here: if this is about the Galaxy Tab 10.1 vs. the iPad (2), why would they cite an RCD from 2004 which doesn't really seem to represent the design of the iPad (2) at all? There's probably some reason for this, but it's a point that most seem to have missed.

The RCD process seems to have obvious issues similar to the U.S. patent process: it is supposed to cover only novel and unique designs, but according to Wikipedia (page hasn't been edited since February, so I don't think there's any 'political' inaccuracy), there is no novelty test at the point of application (and some of the more ridiculous Apple RCDs seem to bear this out). So, like with U.S. patents (where there's supposed to be a novelty test, but it's frequently not done very well), it's probably easy to get an RCD granted which shouldn't really have been granted, and rather harder to invalidate it.

I sorta hate routers, round #2

An exciting and very-nearly-as-good sequel to the classic I hate routers!

After that last little debacle I've been carrying on running the old WRT-310N, but today it finally narked me off sufficiently again. The wireless congestion in my area is just nuts now: I can see 26 wireless networks from my laptop, all in the 2.4GHz band. Even with the best config tweaking I could do, I just couldn't get wireless transfers past 2MB/sec or so - that's a third of the 802.11g nominal transfer rate, never mind 802.11n.

So I did a bit more research, and this time I went out and got a WNDR3700 from Netgear. It's about the highest specced consumer router you can get, and has all the features I was looking for (again): gigabit, dual wireless frequencies, and dd-wrt support. Actual documented dd-wrt support from lots of keen users.

Got it home, plugged it into a single system to configure it, and found I couldn't get to the router's address. At all. Got frustrated for half an hour. Tried a different system, and it worked. Huh. Flashed it to dd-wrt, following all hopping-on-one-leg-wearing-wellington-boots-under-a-full-moon-while-sacrificing-a-chicken router flashing procedure, swapped it out for the WRT-310N, saw everything come up, then noticed it didn't have a WAN connection. As I started to investigate why, noticed that suddenly all network communication completely ground to a halt. Tried power cycling it a few times, and the same thing would happen - no WAN connection, then no connections at all in short order. Tried transferring a file from my laptop to my desktop right after startup, just to see how fast the wireless was - transferred 1MB and then all connections ceased.

Cursed colorfully for a few minutes, then ripped the thing out and put the WRT-310N back in. Entertained wonderful destruction fantasies, but eventually calmed down enough to go trawling through the dd-wrt forums, whereupon I came across this gem (Fedora bug report here). Apparently this particular new-hardware-quirk is not just the universe hating me, but a widely reproduced bug: Intel somehow managed to stuff up their wireless drivers in kernels 2.6.39 and later such that connecting to many dd-wrt router configurations produces exactly the Sudden Death Effect I was hitting. Switched the WNDR3700 back in, booted my laptop back to kernel 2.6.38, and lo and behold, the report's bang on: the local network now works great, and I can actually transfer files from my laptop to my desktop at 10MByte/sec! It's not exactly 300Mbit/sec, but it's a hell of a lot better than a kick in the teeth. Thank you, 5GHz band, thank you.

I still don't have a WAN connection, though. No worries - that turns out to be just a simple case of power cycling the modem. Simples.

So finally I appear to have a working dual-band router setup capable of vaguely decent wireless transfer speeds. Fingers crossed. I highly recommend that 802.11n config tweaking guide I linked to earlier, btw - it has lots of good tips for actually achieving 802.11n connections. Like you have to configure for WPA2 only, with AES encryption only - don't use mixed mode, don't use TKIP. Only use channel 1, 6 or 11, and do a site survey (dd-wrt has this feature) to see what channels other local wireless APs are on. Don't use 40MHz channel width in the 2.4GHz band unless you live in the middle of nowhere, because it's all but guaranteed to be impossible to find enough free spectrum to manage it. But if you have the kind of 2.4GHz congestion I do, nothing's going to help, really. A dual-band router is a really good idea, once you get it running. I've set up a G-only, WPA1/2 mixed, AES/TKIP mixed AP on the 2.4GHz band for compatibility, and an N-only, WPA2-only, AES-only AP on the 5GHz band for security and speed. Definitely seems like the way to go.