A Correction; Welcome to 51,254 (2000) | Hacker News Reader

A Correction; Welcome to 51,254 (2000)

nytimes.com

64 pointshk__214 comments

I don’t think this should have been corrected. Now any reference to an issue number has to be corrected by 500, depending on whether it references an issue before or after the correction.

Other than correcting a number sequence error, this doesn’t fix anything important.

It’s worse: they introduced ambiguous numbers.

(-; They should have done what computers do when they learn their clock runs fast: instead of turning the clock back, they let it run slow for a while. So, No. 51,753 shouldn’t have been followed by No. 51,254, but by No. 51,753.9, then 51,754.8, etc. to reach the right number in 5,000 editions, or 51,753.5 to catch up in 1,000 editions ;-)

That said, it doesn’t break anything important, either, and it does bring the numbering in line with a major newspaper’s understandable insistence on accurate representation of the facts wherever possible. The numbers could quite reasonably be construed as arbitrary identifiers, as you have. But to the Times, they are a statement of fact — i.e., a statement of how many issues have been published. You’re right that it was silly to correct the numbering, but it was corrected as part of adherence to a journalistic culture that exists for un-silly reasons. So I appreciate the correction and find it pretty amusing.

> That said, it doesn’t break anything important

No, they clearly damaged something important - the uniqueness of each issue number.

If the issue number identifying a specific issue is not important, why have an issue number? They may as well label all their editions '1', and save ink. They already have thousands of editions that bear the wrong number; they can't fix that.

Indeed, I am suspicious that someone surely noticed the misprint in the weeks or months following when it was first made, then chose not to fix it for that exact reason.

I take issue with issue number being "important". Each NYT still has a much more useful unique identifier, which is the day it was published.

The issue number is probably entirely useless, but the reason to keep it around are historical. It's a neat thing that they've labeled all their issues like this, right from the start, and that they've kept up the tradition. Like quote from the linked article:

> ''There is something that appeals to me about the way the issue number marks the passage of time across decades and centuries,'' said a memo from Mr. Donovan, who is 24. ''It has been steadily climbing for longer than anyone who has ever glanced at it has been alive. The 19th-century newsboy hawking papers in a snowy Union Square is in some minute way bound by the issue number to the Seattle advertising executive reading the paper with her feet propped up on the desk.''

He's not wrong!

Not everything has to have absolute utility. Not everything has to be perfectly calibrated for storing in a database. Some things we keep around because of tradition, or because we're romantic creatures who sometimes appreciate things which have little practical use.

I think the attitude they're taking is that every edition does have a unique number; it's just that many of those numbers were printed incorrectly.

There are similar ways where fixing a bug introduces new complexity as old data and code were dependent on buggy conditions, requiring to reserialize/convert all old data to newest "FIXED" format or to add an abstract interface that corrects it on the fly.

Do people really reference NYT pieces by issue number? There are actually references like "The front-page article on issue 54,324" instead of "The front-page article on December 4th, 2007"?

The importance of just giving each issue a UUID.

The date itself is already a universally unique identifier (provided there is never more than 1 issue per day).

Also naturally sortable if represented as YYYYMMDD, unlike default guids (unless specifically using sequential guid generator)

Being pedantic here, and somewhat off topic.

In general, a "date" (assuming you mean year, month and day) is not universally unique unless you provide the calendar system under which it occurred.

See: https://en.wikipedia.org/wiki/Dual_dating

This probably doesn't matter much going forward, but it does matter when addressing historical dates.

sha512(top story title + issue number + date)

or, put it in the blockchain! (/s)

A hash of things that could be corrected is way worse than a random number.

Reply on news.ycombinator.com