Slashdot Mirror
Falsehoods Programmers Believe About Names
Jamie points out this interesting article about how hard it is for programmers to get names right. Since software ultimately is used by and for humans, and we humans are pretty tightly linked to our names (whatever the language, spelling, or orthography), this is a big deal. This piece notes some of the ways that names get mishandled, and suggests rules of thumb (in the form of anti-suggestions) to encourage programmers to handle names more gracefully.
Names of what?!
Fuck systemd. Fuck Redhat. Fuck Soylent, too. Wait, scratch the last one.
Software is NOT designed to be perfect and cover every case. Have a numeral in your name? Too bad. Need some names to be case sensitive, and others case insensitive? Sucks to be you. Have a 200 character name that doesn't fit in the 100 characters the designers thought no crazy person would ever have? Tough.
I started reading through the list, and it's just ridiculous. There's a few good points, like names don't change, or names are unique. But they're so obvious that the vast majority of the times it's not a big problem. More often it's just a matter of training the data edit/entry folks how to change someones name, or how to not assume a name is a sole identifier.
But assuming the worst and trying to design a system that'll allow people's names to be Chinese characters when you don't do business in China, have presence in China, or ever ever plan to? That's ridiculous. Software doesn't have to be perfect out of the shoot. It should be adaptable though if some unforeseen shortcoming becomes a larger problem. Gee, I guess if you ever chose to do business in China and need Chinese character names you might have to re-write part of the damn software. Oh well, that's what software developers are FOR!
If you don't even HAVE a name, then I submit you're crazier than the artist formerly known as the artist formerly known as Prince. At least HE had a name, though it was an unpronounceable symbol. The world can't accommodate every possibility, and software is no exception.
AccountKiller
He's essentially arguing that, because names vary a lot and are complex, your software should never do anything useful with them. Sorry, but that's a stupid answer. In a lot of systems, being able to sort by surname may well be more important than being able to handle people who claim they have no surname.
Of course, you shouldn't gratuitously do stupid things, and interfaces should aim to be relatively clear. But most people can figure out how to enter their names into relatively standardized forms, and those that don't should probably figure out how.
10 PRINT CHR$(205.5+RND(1)); : GOTO 10
My last name is O'Leary and over the past 5 years web sites have not gotten any better, and arguably have gotten worse, at handling the apostrophe in my last name
Help me Slashdot, you're my only hope.
First thing I learned back in 1993 when I got started.
1) George Foreman has five boys named George Foreman. Your database better be able to handle that. :-)
2) Your database better be able to handle Cher (no last name).
3) People are not required to have Social Security numbers. (it's an optional program - you don't have to partipate).
4) Not everyone's last name starts with a capital letter.
5) Mexican people's names break ASCII (the tilda n).
6) People named O'Grady have a hard time getting their name in a database sometimes and have a hard time getting their name passed via a URL sometimes and generally mess stuff up.
7) People from Sri Lanka will break your name length limits.
8) Some people's name is only a single letter.
9) Some people go by their middle name god damn it!
I code to spec. The product and marketing departments write the spec (what little there is); the QA department amends the spec with overly specific test cases. I suggest that the spec is incomplete and won't handle...but I'm told, just code it to spec. I recommend changed, but we don't have time for edge cases. I point out potential problems, but we're unlikely to get any of those. I warn of potential compatibility problems but we don't care. Are you just trying to be difficult? If there's a problem QA will catch it. The project is overdue already, and by the way here are some new requirements that need to make it in, and we can't change the release date because we already promised the stockholders. Why is your code so complicated, my twelve-year-old kid could write this.
It's not my fault. I code to spec.
It seems to me that most misconceptions about names can be fixed by the following:
Allow a single, Unicode-enabled field of "unlimited" length (let's say 4 kilobytes) which represents "name". Several would be defined by different roles -- "Real name", "Nickname", "login", where only login (sometimes simply an email address) is required to be globally unique.
Now let's look at what that breaks:
First, #1, 2, 4, and 5. How am I supposed to avoid assuming these? People should be allowed to enter an arbitrary number of names for themselves? I suppose that's possible, but it immediately kills most of the potential uses of this data. If I want to set a nickname that goes with my forum posts, say, what good is it for me to have five nicknames? Seems like the only potential use would be making people easy to find by real name -- so, a social network.
#6 -- surely 4k is enough, but this is also not a terribly difficult assumption to change later. Annoying, but not devastating, not even as hard as changing from the first name / last name combination into one "real name" field.
#7, 8 -- most systems would make it trivial for people to change their names.
#9, 10 -- UTF8 is easy.
#11 -- very, very curious to see an example. And wouldn't that be a bug in Unicode? And this is again one where I have to ask -- how do you change this? Allow arbitrary images?
#12, 13 -- obvious solution is to make the name system case-preserving, thus allowing both case-sensitive and case-insensitive searches.
#14 -- again, avoid by simply allowing the name to be a single opaque field.
#15, 16, 17 -- if your name supports random unicode, no idea why these would be a problem.
#18 -- not sure why it matters.
#19, 20 -- again, if it's just arbitrary text, it just works.
#21, 22, 23 -- not sure how I'd make that assumption.
#24, 25, 26, 27 -- again, the name is just an opaque bunch of characters.
#28 -- what?
#29 -- opaque characters.
#30 -- keep the original text as-is. If you want to try to split people out by naming scheme, do it later, but keep the original. This should be a "duh" concept -- always preserve the original user input. Cache transformations for speed, if you like, but they're a cache -- keep the original. Your algorithm might change.
#31 -- bad idea to assume bad words won't cause problems in general. I currently play an MMO in which I physically can't talk about Emily Dickinson, and have occasion to more frequently than you might suspect.
#32-36 -- why would it matter? Unless...
#37 -- Fine, but how would I otherwise connect the same person?
#38 -- How about unicode-equivalent? And of course, they might not -- one might make a mistake, or the name might be represented differently. But you'd have to deal with typos anyway, so this isn't exactly shocking.
#39 -- I'm going to have to agree with the assumption, though. If I develop a system which works well for people who only follow the US standard, and I suddenly have a ton of people from China wanting to use my service -- enough that this is actually a problem for me -- that's a nice problem to have.
#40 -- People can make up names. I guess this explains #32-36, though.
The sense I get is that half the list is stuff you'd almost have to be stupid to run into (seriously, who doesn't use Unicode?), and the other half involves some seriously weird names and cultures that are going to have to meet me halfway, if they expect me to do anything interesting with their name. As I understand it, the only way to get this right would be to allow people to have zero or more names, each of which is either an unlimited amount of text in any encoding, or an image (raster or vector) of unlimited size. To query such a system requires insane amounts of logic just to deal with the text, and throw in some OCR for good measure.
I think this is a case where I would much rather see people evolve to match the technology, rather than the other way
Don't thank God, thank a doctor!
The regular expression, if one must be used, doesn't need to be any more complex than:
^[^@]+@[^@]+$
Sending out response emails to an improperly validated address just turned you into an open relay. Spammers can use your server to send spam by embedding their entire message as the email address, trailed by '\x004@.'
Validate your inputs. Always.
Sometimes I despair when I read or hear somebody referring to eg. Djengis Khan as "Mr Khan" ("Khan" is a title, not a name) or even call Hu Jintao, "Mr Jintao"; you would have thought people would, by now, have caught on to the idea that something like half the world's population has the family name first.
Oh, come now - are you seriously saying you expect every single person to understand every subtle nuance of every other culture's use of titles and names? Here are some non-English equivalents to Mr., are you seriously telling us you know all of these? Here are the various forms of address in the UK alone, do you know all of these and every other culture's equivalent? How many of these should I learn before I go from being someone you despair of to someone you feel is welcome in your titular elite?
If half the world's population has the family name first, which half do I choose to offend when I don't know the exact rule for the home country of the person I'm speaking to? That's even assuming I know which country they're from. There's no reason to assume in this shrinking planet that someone who looks like they're from country A wasn't in fact born in country B to parents from countries A and C - a person born in Japan but with lineage in China might take great offence if I use Chinese honorifics to address him, surely it's better to be polite within the confines of my own known culture than to make such crass assumptions about his? The key thing I take from someone saying "Mr Khan" or "Mr Jintao" is that they're at least making the effort to communicate in a civil manner, which certainly causes me no despair.
You know, attitudes like yours are IMHO the root of all that's wrong with computers today. And I'm saying that as a programmer, not as Jane Grandma. The whole idiotic OCD idea that you _must_ make up rules about everything, and that your rules are more important than what people are actually trying to do. The idea that if even someone's name doesn't fit "your" database, then you can just brush them off and have a beer.
Here's some free clue: yes, you can't handle every edge case in the universe, but you'll find it's easier if you don't create such edge cases in the first place. If your database (actually more likely the program in front of it) can't handle last names with more than one capital letter, or with a dash in the middle, or which are more than 32 bytes long (which with UTF-8 might mean less than you'd think), then guess what? _You_ created an artificial edge case that had no reason to be there in the first place. Instead of handling every edge case in the universe, how about not creating them in the first place?
I find that about 90% of the problems boil down to the above: some idiot put some artificial limits or rules, that really aren't needed anywhere else. Just because he has the delusion that he's some kind of Moses on the mountain and just _has_ to come down with some rules.
E.g., he just had to define a byte limit, because he's prematurely optimizing a non-problem he doesn't understand. God forbid wasting space in the database by allowing 256 or 2000 byte strings... never mind that if he actually understood that underlying database, he'd know that a VARCHAR is not padded to that max length. If someone just entered "Alex", the same 4 bytes will be actually used in the database, regardless if the field is a defined as maximum 4, 32, 256 or 2000 characters. But nah, he has to put some restrictive number there, 'cause it looks more like he's doing some smart job.
There is hardly any reason to even use a user name for anything other than display purposes. (You do have a primary key for that record for everything else, right?) As such there is no reason to make any assumptions about it, or enforce any particular format, or anything. There's no reason to even disallow SQL keywords (just effing quote it before using it in SQL) or angular brackets (just quote it before using it in HTML.)
There is no reason to create any edge cases in the first place.
And really it's not even just about names. Names are just one case where people make up BS rules just to feel more like they did the great design job. One could make the same case for the gazillion other pointless rules imposed upon the user or his work-flow or data, not because they're actually needed anywhere, but just because some OCD idiot feels like he _must_ impose some rigid structure upon things that really have none and don't need any. But he'd just feel naked without defining that kind of rigid structure, or without imposing upon humans some data structures theory that was intended only for use by programs.
A polar bear is a cartesian bear after a coordinate transform.