Once again, this useless obsession with “features.” Linux advocates never come up with actual use cases for their “innovations,” because they never consider the user in the first place. The classic instance is, as you say, the Ubuntu “Cloud” drivel, which appears to me to be a very, very expensive way to acquire an off-site backup ($30 a month for 400GB, unless there’s a bulk discount of some sort) which might disappear at any time. The actual “useful” feature, which is essentially a roaming profile, is flat-out impossible on Linux because no two machines behave the same.

I like the plaudits for (Oracle’s, hem hem) Btrfs, which I believe currently has defragmentation (oh, the humanity) but:

“As of January 2012, the planned filesystem check program had not been released. This means that a btrfs filesystem can become corrupt and lose all its files if a machine crashes or loses power on disks that don’t handle flush requests correctly.”

So, the only thing that a user will ever notice about their file system (that it recovers after a hard failure) is, um, not actually there, five years — or to put the timeframe in the terms of the poster, “years ago,” as though 2007 was some sort of geological age — from initial implementation.

Way to go, chaps.

So, the only thing that a user will ever notice about their file system (that it recovers after a hard failure) is, um, not actually there, five years — or to put the timeframe in the terms of the poster, “years ago,” as though 2007 was some sort of geological age — from initial implementation.

Remains me of this

http://arstechnica.com/business/news/2012/02/libreoffice-35-released-introduces-new-grammar-tool.ars

You set the time span for this.

LibreOffice (2011)
OpenOffice (2002)
StarOffice (????)

I notice that Sandro Villinger, in one of his links¹, throws out this headline:

10. Programs and drivers won’t run

... immediately followed by:

“I think it’s nothing short of a miracle that nearly all of my Windows 7 applications and drivers work right out of the gate. Microsoft’s legacy support is spot on, even in this early pre-beta release of Windows 8.”

Unbiased reporting at its finest.

[1] http://www.itworld.com/software/224533/windows-8-10-biggest-problems-so-far?page=0,3

That Grammar 3.5 thing? Dear Lord:

“The feature is built with Lightproof, a language-neutral grammatical analysis tool that is implemented in Python and has a sophisticated regex-based rule system.”

WTF is this? A tool designed to make us all write like oiaohm?

Damn. I had other things to do this afternoon, but now I’m going to have to look into it…

Well, that didn’t take long. The thing is practically invisible on the Web.

Apparently it was conceived of by a mad Hungarian (is there any other kind? I love Hungarians) in 2009 for, ahem, Open Office. For a taster, see this¹ link.

The word “grammar” doesn’t convey the awesome power of this regexp-based beast. Take this example:

“# silent h
(?i)\ba
(honest(y|ly)?|hour(ly|glass)?|honou?r(abl[ey]|ed|ing|ifics?|s)|heir(less|loom)?)\b
-> an \1 # Did you mean:
TEST: A heirloom -> An heirloom”

Two things here. There’s a total absence of stemming, a la Porter, which is a useful technique that has been around for 20 years or more and even has its own Snobolly language (google Snowball for details).

There’s no obvious use of n-grams, which I would have thought would be absolutely essential to a context-based grammar checker. Oops, that wasn’t supposed to be one of the two things…

And, crucially, does anybody seriously believe that the approach in this case is remotely scalable? As an exercise, spend ten seconds trying to count the number of aspirate-free words in the English language beginning with the letter 'h’. Now another ten seconds. And another.

Now then, are we all settled down and breathing properly again, class? Can anybody explain to me why this issue, which is not scalable across a single regexp rule, is going to be scalable across hundreds of non-orthogonal ones?

You can? Have a lollipop!

[1] http://www.mail-archive.com/dev@lingucomponent.openoffice.org/msg02170.html

By “useful” I should be more honest and admit that “useful” only applies to Indo-European languages (and, interestingly enough, several West African ones, with modifications). Not for Chinese, and I doubt for Korean and Japanese and Vietnamese and Thai, although I’d be completely out of my depth on these.

From my perspective, working on Related Search at Bing over the last couple of months (you didn’t think I just pulled these observations out of my awesome depth of universal knowledge of everything, did you?), this is insanely simplistic. Bing hasn’t even got it right yet, and that’s just for the major markets, which tend to be Indo-European. I doubt Google has, either.

One bloke in a bungalow off Lake Balaton or a third-storey flat in Pest is hardly going to be able to beat this massed expertise.

“WTF is this? A tool designed to make us all write like oiaohm?”

That remark, dear sir, made my day.

My s.o. studies computer linguistics, and I’m starting to think developing an oiaohm-bot might be an interesting project for her thesis, seeing as she’s looking for a challenge.

““WTF is this? A tool designed to make us all write like oiaohm?”

Nope if that were the case, it would have to translate into Old English ;-)