LWN: Comments on "Sunday's kernel releases"

Sunday's kernel releases

Wol — Sat, 03 Nov 2012 00:59:45 +0000

Sounds like the senior management at the company I worked for a while back :-)

David Clark the MD, David Wheatley the Operations Director, David ... the Company Secretary, and David Geldart the IT manager.

Cheers,
Wol

usernames

daglwn — Thu, 01 Nov 2012 16:44:05 +0000

A friend once set up a university e-mail list called "hi."

Much entertainment ensued.

Sunday's kernel releases

dlang — Thu, 01 Nov 2012 06:34:12 +0000

I had 4 Davids in a class of 12 at one point.

At work a few years ago we had the 'david meeting'

from Security
Me (David)
my boss (David)
his boss (Scott)
his boss (David)

From Engineering David

From Operations David
and his boss David

Scott opened the meeting by asking if he should change his name to David for the duration of the meeting :-)

Sunday's kernel releases

Wol — Wed, 31 Oct 2012 23:31:43 +0000

In my first proper job, we had three Michaels in a department of six. Fortunately they went by the shortened names of "Big Mike" (the boss), Mike, and Mick.

Then in another job, we had four Anthonys in a company of about 60. So we had two Tonys, an Anthony (me), and a Stiffy (his surname was Stiff).

And my boss there was Brian Smith. Apparently, when he was in short trousers, there were four Brian Smiths in his class (of about 30 kids).

Cheers,
Wol

Sunday's kernel releases

Wol — Wed, 31 Oct 2012 23:27:45 +0000

Well, when my wife was a girl they had great fun at home. She, her elder brother, and her dad all shared the same first initial. When a letter arrived, it was often a puzzle (until it was opened) who it was addressed to.

Cheers,
Wol

Sunday's kernel releases

Wol — Wed, 31 Oct 2012 23:25:41 +0000

For a well known example, Mr Evelyn Waugh married a lady called Evelyn ...

Cheers,
Wol

Sunday's kernel releases

marcH — Wed, 31 Oct 2012 09:06:43 +0000

http://xkcd.com/327/

[This is even more digression but this whole page is just noise anyway, so]

While this cartoon is very funny it misses the technical point. A decent database language would allow garbage-in/garbage-out and not require the insane amount of (brittle) sanitizing that SQL requires.

Prepared Statements solve this problem. While their use should be mandatory they're not even part of plain SQL.

SQL is a joke from another age.

Sunday's kernel releases

bosyber — Tue, 30 Oct 2012 11:11:33 +0000

Twins don't need to marry to get the effect, just having their initials match gives at least quite a problem with some Dutch banks: Postbank, now taken over by ING - which only compounded the issue when one moves.

usernames

jezuch — Tue, 30 Oct 2012 10:32:45 +0000

> Erm... Ruth is a very common English first name

Slight correction: it's common among English-speaking peoples, but the name apparently is Semitic.

usernames

ghane — Tue, 30 Oct 2012 09:29:34 +0000

"Late godfather". Quite sure we are not talking of Enoch here.

--
Sanjeev

usernames

tialaramex — Tue, 30 Oct 2012 02:51:12 +0000

My goodness.

You somehow thought that while I was aware of the existence of a fairly obscure English word, I hadn't noticed it's also a popular name?

I chose the username purposefully. The account which it corresponded to was not of great importance to me (I had other accounts with which to get real work done), so its value as a lesson and example more than compensated for the trouble it caused.

The argument that "it wouldn't be short-sighted to assume that it was a person's name" mistakes the problem. It /would/ be short-sighted to assume that blindly addressing email to "ruth" in an organisation of several thousand people will have any particular effect at all, still less that it will ensure your confidential missive is received by whichever person named "Ruth" you happened to be thinking of when composing it. It would also be short-sighted to create a system which "helpfully" redirects email sent to, say, "Ruth.Smith@University" to a user named ruth on the basis that the email address didn't match anything and maybe this "ruth" account will know what to do with it. It would be even more short-sighted to invent a system which attempts to match named employees to accounts based on the similarity of the username when a perfectly good system for exactly matching corresponding accounts by payroll number already exists. And still more short sighted to write software which treats the human readable part of an email addresss (e.g. the Ruth in "Ruth Stevens" <stevensr@University>) as a username to which the mail should be delivered. And yet all this short-sightedness and more happened.

Sunday's kernel releases

anselm — Tue, 30 Oct 2012 00:08:19 +0000

For extra credit, marry a woman with the same first name and the same date of birth.

Sunday's kernel releases

NAR — Mon, 29 Oct 2012 19:23:29 +0000

When I was a kid, we used to play football (that's soccer for some of you) with neighbouring kids. Teams of 5-7 boys, depending on who was around. I remember one case when we used the following convention to select teams: everyone called Attila plays one team, the rest are in the other. We won :-)

usernames

Cyberax — Mon, 29 Oct 2012 18:03:04 +0000

Is his first name Enoch, by any chance?

(for those missing the context, read Neal Stephenson's "Cryptonomicon")

usernames

Cyberax — Mon, 29 Oct 2012 18:01:00 +0000

>You can't even GUESS at their username, because it's numeric

Tattoo it on their foreheads. Make it a barcode for easy machine readability.

Duh, that one was easy.

Sunday's kernel releases

shaworth — Mon, 29 Oct 2012 17:46:12 +0000

Try marrying a woman with the same first name. The result, identical first and last names, addresses and phone numbers.

Namespace collision, more common than you think.

Mr. Shannon Haworth

usernames

deater — Mon, 29 Oct 2012 14:51:15 +0000

my late godfather's surname was "root". I doubt he ever had an account on a UNIX machine, but I've always wondered what kind of troubles he could have caused if he did.

usernames

ledow — Mon, 29 Oct 2012 13:31:31 +0000

Now imagine you have a class of 30 6-to-11-year-olds, all with unique numbers, all waiting to log into the ICT Suite. It's their first lesson, you're the teacher... see how much you get done.

You can't even GUESS at their username, because it's numeric. They won't remember it, even months later. You will have to print little cards with their usernames on, which they will be unable to read until years after they've started (BY LAW!) using computers in a network environment so the teacher has to go around one-by-one logging them in.

Not saying it's insurmountable (it's not - for a start, I push for dongle-logins because it's just easier even if they lose them in the first five minutes), but it's a problem.

Hence why humans memorise alphabetised mnemonics (i.e. a username based on their name) and let the computer convert it to a unique identifier (i.e. user-SID or equivalent).

The only place that gives me a numeric username is the UK's Government Gateway (which also needs several highly-secure passwords and security procedures to access because it lets you do everything from file your taxes to renew your driving licence). And that's been phased out because they get so much hassle with people forgetting their logins and they're planning to tie it into email addresses or social network accounts or similar.

We use usernames for a reason. If I wanted to use numbers, I'd just allocate them a login-dongle of some kind. Good luck tracking down user123127854738's history on all your systems, even if you *do* implement logging and searching. And username auditing (is this account still in use?), and lots of other boring admin tasks which are solved by using some variation of real name even despite the inherent namespace problem.

usernames

ledow — Mon, 29 Oct 2012 13:21:43 +0000

Erm... Ruth is a very common English first name (and slightly less common surname).

My boss at a previous school was called Ruth, for instance. I know someone in American called Ruth (though she goes by a nickname because Ruth is seen as old-fashioned). And then there's Babe Ruth and other famous "Ruth"s, first or last name. It wouldn't be short-sighted to assume it was a person's name at all. And that's exactly my point. Any system of naming will run into a name from some other place sooner or later.

usernames

sorpigal — Mon, 29 Oct 2012 13:19:21 +0000

Everyone has this problem and the solution is simple: Guarantee uniqueness by assigning every person a unique number. The problems with this are that people feel numbers to be impersonal (as if most of these schemes are not!) and that numbers can be harder to remember or distinguish from other numbers. Attempts to avoid this problem with clever schemes will, as you point out, lead to trouble[0].

There are no good answers, but here's a bad one that I like: assign each person a name-based username and then *always append a numeric suffix*, so that each name is e.g. firstname.lastname.123@host. Of course you have to deal with the fact that nobody really knows what a name is[1], allow for exceptions and never hard-code this assumption in to any software.

Perhaps the best way to name people is to use something like DNS. A unique number that identifies plus a global name-to-number resolution system, which is non-authoritative and permitted to change over time, geography, etc.. Of course, then you'd need to know he point in time and locality to know how to resolve the name... which is pretty much where we are now.

[0]: http://thedailywtf.com/Articles/The-Automated-Curse-Gener...
[1]: http://www.kalzumeus.com/2010/06/17/falsehoods-programmer...

usernames

tialaramex — Mon, 29 Oct 2012 12:13:48 +0000

Let users pick their own username, forbid already existing and recently deleted users, laugh at anybody foolish enough to insist on their forename as they are constantly mistaken for somebody else with the same forename.

Since they are picking the username, they are solely responsible for its perceived meaning if any. If you're dealing with kids I suppose you might thus want to have oversight/ veto.

If absolutely necessary also forbid usernames which conflict with system identifiers, but prefer where possible to simply separate the two namespaces altogether so that it's impossible for either the system or the users to confuse a person with a machine, or an ordinary user with an administrator or other higher power.

While working at a university I selected the username "ruth" (a now rare word meaning roughly "remorse" and the source of the still common word "ruthless") and the associated email address was assigned to me automatically. Over the years I discovered that humans who can't use their email client properly are about as common as automatic systems that are too dumb to know the difference between a username and a person's real name. I received confidential correspondence, invitations, enquiries, bounces, and numerous other emails that should have been sent to (and in some cases, explicitly were addressed to) members of staff with the forename "Ruth". Nobody showed any interest in either fixing the automated systems or retraining the staff, but they did eventually institute a system in which they gave everybody an email address of the form Firstname.Lastname@University. This system had all the flaws you have mentioned above, it fixed nothing, but most likely it made some barely computer literate new VC feel that they could email anybody by guessing at the correct address and then probably blaming "those IT people" when their confidential union settlement proposal was mistakenly sent to a janitor or whatever.

Linus needs to get used to this, or get out!

linusw — Mon, 29 Oct 2012 11:30:34 +0000

Torvalds is actually just three years older than me (Linus Walleij). And both of us were probably inspired by proxy from Nobel laureate Linus Pauling.

Sunday's kernel releases

jengelh — Mon, 29 Oct 2012 11:23:42 +0000

That's just what I linked to, is it not?

Sunday's kernel releases

mpr22 — Mon, 29 Oct 2012 11:07:27 +0000

Initials + serial number should let you fully automate filtering for the likes of Simon Henry Ian Thorpe and George Oliver Donaldson.

Linus needs to get used to this, or get out!

k3ninho — Mon, 29 Oct 2012 10:56:33 +0000

Nominative determinism calls for it - though I suspect a large number of prospective kernel hackers called Linus have the ball pulled from their toes just as they kick at it.

Sunday's kernel releases

ledow — Mon, 29 Oct 2012 10:53:07 +0000

I know someone whose dad is David, whose brother is David, whose married to a David (who has several friends and relatives called David) and whose boss is David.

Also, working in schools, I can assure you that ANY naming scheme you choose for users will inevitably invite a double-entry because of the way probability works - even in a school of only a few dozen students.

And where you come up with a smart-alec user naming scheme, there will always be doubles or you'll have to add a number to the end because of exactly the same problem. Every single place I've ever worked has had a duplicate username identified by, say, "username2" or similar, and the same problem with every influx of new users.

And, worse, if you use a smart-alec scheme like "first four letters of surname followed by first two of first name", you will invariably end up with something approaching a rude-word and it will ALWAYS be the most senior person in the place (so, obviously, you are then asked to explain and to change the naming scheme or provide an exception which breaks all your user scripts). I actually have to audit pupil names for pronunciation and possible connotations before I issue users (very easy to get swear words, insults and I guarantee you the pupil with the most fanatically religious parents will end up with something blasphemous under the naming scheme).

Sunday's kernel releases

magfr — Mon, 29 Oct 2012 06:25:19 +0000

Something along the line of this I suppose?

Sunday's kernel releases

jengelh — Mon, 29 Oct 2012 00:15:43 +0000

Don't know what to name your newborn? Forget books on names and etymology — just pick something random from `git log --format=pretty:%an | pcregrep -o '^\S+' | sort -u`!

… Just don't pick something with “=?utf-8?q?” in it.

Linus needs to get used to this, or get out!

charlieb — Sun, 28 Oct 2012 23:50:26 +0000

These could be the vanguard - 2nd generation linux hackers. You think none of them will be named Linus? :-)

#15091 here

man_ls — Sun, 28 Oct 2012 23:26:57 +0000

I sympathize, having being born a "Fernández" in Spain. It feels good to also carry a prime identification number, even if there are infinite of them. And an irregular prime too!

Sunday's kernel releases

neilbrown — Sun, 28 Oct 2012 22:54:40 +0000

Hi,
I am #359. Having been born a "Brown", it is such a relief to be unique (and prime too!) in this little community.

Sunday's kernel releases

nix — Sun, 28 Oct 2012 22:48:57 +0000

It could be much worse.

Your name could be Steve.

Sunday's kernel releases

mikemol — Sun, 28 Oct 2012 22:44:36 +0000

My name is Michael. Among my best friends are another Michael, two Joshes and three Davids...and one of those Davids goes by the name Myke.

I have absolutely no sympathy. :)