Hacker News Reader

Comment by ademarre | Hacker News Reader

When it comes to censoring randomly generated strings, I like simply to omit vowels from the alphabet. Usually I'll omit some of the more obvious lookalikes too, e.g. [1 0 v].

It's a simple solution. Sure, it is still possible for something to slip through that looks similar to something bad. But the potential to strongly offend is greatly reduced.

View on HN

Yeah there was an article I read a while back about a company looking to prevent the use of 'naughty' words in randomly-generated strings used as event IDs. Apparently someone with some pull had seen a message with an offensive word. Some management committee spent a long time trying to figure out how to solve the problem including proposing keeping a list of bad words, and then worrying about what should be in it and who would maintain it. At some point an engineer got a chance to speak and said something like "just use base-31 and omit vowels". The story as I remember it didn't mention the use of v or l33t-speak, but they were randomly generated, not maliciously constructed, values.

Also note that if you're too naive about checking for 'naughty' words, you get https://en.wikipedia.org/wiki/Scunthorpe_problem

> didn't mention the use of v or l33t-speak

What is 'v' in this context?

Edit: thanks for the answers. It makes sense now.

I suppose it's about the use of 'v' instead of 'u' in the F word

It's visually equivalent to a u, so without vowels it's still possible to get "fvck", for example.

The Romans wovld approve of svch a scheme.

How would the Romans have pronounced 'approve', though?

imagine the phrase "see you later" was offensive, or perhaps its abbreviation "cu"

the combo "cv" could then become problematic.

Hashids does this (avoid bad words) if anyone is curious to see an implementation

> algorithm tries to avoid generating most common English curse words by never placing the following letters (and their uppercase equivalents) next to each other:

> c, s, f, h, u, i, t

https://hashids.org/#how-does-it-work

E: ah it was already mentioned later on, hadn't got that deep into the comments yet!

Oh yes, there's plenty of ways to avoid curse words, but each one has a cost, and if the system needs to generate ids very quickly, any scheme that works too hard could be a bottleneck. The naive way of just randomly throwing together letters of the alphabet will eventually generate a forbidden word. Any steps taken to reduce the probability should understand the time/space tradeoffs.

Gov.uk did this. The code WNKR still ended up on Reddit yesterday.

That's amusing, but I think it also highlights the effectiveness of the strategy. WNKR is excusable and defensible. WANK would not be.

Edit: But I'll concede that when your outputs are only four characters long and end users will actively interact with them (write them down, type them again later, etc.), additional safeguards might be appropriate. Or simply omit all alphas and use only numerics.

> Or simply omit all alphas and use only numerics.

You're still not out of the park with numerics - people with 1313 or 6660 or 4444 or something will complain a lot. The possibility of a 666 in some new biometric government IDs in my country rose a massive stink from church...

I was in the UK recently and - if you can believe it - there was a car on the block I stayed whose plates contained 666.

Also, have a feelin you meant to do 1312. What’s the issue with 4444, though?

> What’s the issue with 4444, though?

https://en.wikipedia.org/wiki/Tetraphobia

Woah, had no idea!

> When Beijing lost its bid to stage the 2000 Olympic Games, it was speculated that the reason China did not pursue a bid for the following 2004 Games was due to the unpopularity of the number 4 in China. Instead, the city waited another four years, and would eventually host the 2008 Olympic Games, the number eight being a lucky number in Chinese culture.

Thought this was particularly interesting.

In Germany, where you can request number plate combinations (as long as they are free and follow a few roles), 666 is a pretty common combination amongst young drivers.

> What’s the issue with 4444

4 is pronounced similar to "death" in sino-japanese languages and dialects.

Also, as you have the muncipality-shortcut at the beginning, and then two user-definable letters, you freqently see "rude" combinations, and noone bats an eye. BIT-CH, MON-GO, ANA-L, DIL-DO. (And those are just the combinations that are understandable in english which I saw when driving). I'll never forget the face of the guy at the Zulassungsstelle when I was there with a friend who wanted COC-K-6969. He got it it, btw. When I tried some years later, AC-DC-666 sadly was already taken.

What’s the issue with 1312?

it spells ACAB if you match each number with the letter in the alphabet at this index (I realized that after seeing a bunch of 1312 tags around where I live)

What's the issue with ACAB?

https://en.wikipedia.org/wiki/ACAB

People explained 4444, with 1313 I was thinking of the equivalent for superstitious westerners afraid of the number 13.

> biometric government IDs

Yeah. An important, long-lived ID that will stick with an individual for their entire life, and that they may want to commit to memory. That seems like a good time to take a hypersensitive approach and adopt some kind of filter.

My girlfriend got a new bank account and when she received her account number it contained 666. She asked for a different number and they changed it without charge.

> Or simply omit all alphas and use only numerics.

That works pretty well until you realize that some numerical combinations are common neo-nazi codes and may lead to ... unfortunate associations. The ADL lists a few of those^1, but the list is by far not comprehensive, codes actually differ based on locality, and accidental combinatory collision in a 10-character space than it is in an alphanumerical 36-character space.

[1] https://www.adl.org/education/references/hate-symbols/88

Removing wovels would increase likelyhood of bad word combos, but removing consonants would have the desired effect.

Reply on news.ycombinator.com