[X]My Assistant

[WildBird]

Parents usually name twins differently to disambiguate, so your example is an extreme rarity.

Haven't you seen The Newhart show? Hi, I'm Larry, this is my brother Darrel, and my other brother, Darrel.

And I wouldn't say extreme rarity, there could be a lot who have the same first initial. I worked overseas in a few developing countries. They often don't know their birthdays, just maybe year and month.

To me, a good design caters to the lowest common denominator if possible.

[FrankRuperto]

WildBird,

I agree, I have been overseas where tons of people have identical last and first names, names like "Harish Patel", "Shamin Begum", "John Smith", "Mario Di Roma", etc.
In a Crow Indian Reservation close to where I used to live in Billings Montana USA, almost everyone's last names were "Moccasin", "Yellowtail", "Curley".
Parents in the old days named their children with a given name and a surname describing where they're from, the family's trade/skill, etc. and these names have propagated throughout generations to present day.
So we can't rely on just surname's and given names, rather we have to gather as many attributes possible from each person and use a combination of all that to build a unique profile.
This post has been edited by FrankRuperto: Dec 2 2019, 07:14 PM

[WildBird]

Frank,

We all know names are in no way unique. I myself work sitting next to someone with same first name, and a few others in the immediate vicinity, so I get Coop, not Stephen. Its just white noise now when I hear Stephen.

What I have issues with is people thinking fingerprints or DNA are in fact unique. While I agree they can be used to differentiate i.e. you can say that they are different, I don't believe it is possible to say they are unique.

Take a simple example. You have a database of 100 million customer records. You are looking for a Bob Smith (assume we search all variations of Bob, Robert, Bobby, etc.), born 1st January, 2000. A simple SQL search finds 1. Does this mean that there is only 1 Bob Smith, born 1st January, 2000? In this dataset, yes. But with 100 million customers, there must be some history. So you look at archive tables, to expand the data set. Search still only turns up 1. Does this mean there is only 1 Bob Smith, born 1st January, 2000? Well no, because he may not have been a customer. So even if you looked at all the providers in the country, and in their archive tables, there might not be another Bob Smith in the country. But what about overseas? Even if you looked at every provider, and every archive table, there is still a chance that there was another Bob Smith, born on the same day, but never was a customer anywhere in the world.

My point is, that comparing fingerprints or DNA, this is simply a number of fields, more than the simple example of names and date of birth, against a known dataset. The issue is that it is impossible to compare against every single person, living or dead. If we are talking about a crime and literally life or death, then I can only see DNA and fingerprints being able to be used to say 2 samples DONT match.

I also feel we have moved away, ever so subtly, from the original post :-)

[nvogel]

Is this even relevant to database design or data management? It's not a problem because hardly anyone is interested in identifying customers by DNA fingerprints. Millions of people and their database systems work just fine with other identification schemes because they are perfectly adequate for their purpose.

[gemmathehusky]

I think you can take it for granted that fingerprints are unique. The chance that 2 different people with the same name have the same fingerprints but are different people are vanishingly small.

If all this wasn't the case, then replication keys wouldn't work either. Access (other databases?) are predicated on the fact that a random numeric key won't be duplicated, and there are only 2 billion long integers. A database for every person in the US with a population of 300m, might start getting random hits for key numbers only, though. You would be getting serious issues with China, and a population of 1.4billion.

I assume a bigint uses an 8-byte integer, not a 4-byte integer.


@nvogel
With regard to data systems being "adequate" - see this example in the UK
https://www.bbc.co.UK/news/UK-politics-eu-r...rendum-35959949

reported "immigration" is a third of the number of NI (ie SSN) numbers requested.

I wonder if there are stats available for the US. Apparently you have used about half the capacity of SSN numbers available.

[FrankRuperto]

If fingerprints are good enough evidence for a court to decide if a person is guilty or not, then they are reliable enough to uniquely identify someone in a database
So getting back to the OP's question, using just LastName and FirstName is not going to be good enough to prevent duplicates and the OP will have to periodically check for them.

P.S. Dave, that BBC link you posted is 404.
This post has been edited by FrankRuperto: Dec 3 2019, 08:24 AM