3 ms·

I see some comments conflating privacy of sequence statistics with global uniqueness considerations and UX. If your concern is globally unique identifiers (i.e

I see some comments conflating privacy of sequence statistics with global uniqueness considerations and UX. If your concern is globally unique identifiers (i.e. so that you can merge tables across multiple instances of your database), then UUID is exactly what you want. This is entirely what it is designed for. If your concern is the privacy of sequence statistics, then UUID incidentally solves your problem. It may not be precisely what you want, and could continue to leak private information depending on the specific variant used. If you want privacy of sequence statistics, then I would suggest something like a sha256 hash of the primary key concatenated with a cryptographic salt stored in a separate column. These make excellent identifiers in places like APIs and URLs. If you desire a unique identifier that has a high quality UX, then this is in addition to the above columns. This sequence generally has lower entropy than the cryptographic approach (or the UX would suffer), so additional measures should be taken to protect the privacy of the identifiers (e.g. expire redemption tokens after a period of time and reissue via email). Autoincrementing integers are really nice if you don't actually need UUIDs. Lots of tricks you can apply with them.

almog 2y ago
> It may not be precisely what you want, and could continue to leak private information depending on the specific variant used. If you want privacy of sequence statistics, then I would suggest something like a sha256 hash of the primary key concatenated with a cryptographic salt stored in a separate column. These make excellent identifiers in places like APIs and URLs. Is there a reason not to use version 4 UUID and if time ordering is needed, save timestamp explicitly as another column?
- bob1029 2y ago
  Type 4 guids are a reasonable choice. I personally prefer the extra 128 bits of entropy, but I have no evidence that says it's more secure in practice.
tinodb 2y ago
Most of the comments here seem to gear towards “you must always…”. But as you rightfully point out, “it depends” and is an architectural trade off depending our the required qualities.

almog 2y ago

> It may not be precisely what you want, and could continue to leak private information depending on the specific variant used. If you want privacy of sequence statistics, then I would suggest something like a sha256 hash of the primary key concatenated with a cryptographic salt stored in a separate column. These make excellent identifiers in places like APIs and URLs. Is there a reason not to use version 4 UUID and if time ordering is needed, save timestamp explicitly as another column?

bob1029 2y ago
Type 4 guids are a reasonable choice. I personally prefer the extra 128 bits of entropy, but I have no evidence that says it's more secure in practice.

tinodb 2y ago

Most of the comments here seem to gear towards “you must always…”. But as you rightfully point out, “it depends” and is an architectural trade off depending our the required qualities.