From the internet, it seems that GUIDs are kinds of 16 bytes hashes and so are meant to be pretty unique and are generated when partitions are created.
However, sites like Wikipedia (see https://en.wikipedia.org/wiki/GUID_Partition_Table, chapter "Partition type GUIDs") reference some partition type GUIDs. Indeed, my SWAP partition GUID (which I can obtain via "lsblk -o parttype') is the same than the one on Wikipedia.
So, are GUIDs really unique please?

You muddle up GUIDs used for identificating the partition itself and the type of partition. GUIDs of partitions itself, of course, have to be unique. But GUIDs used for types are not generated by partition managing software every time: they were generated once when the standards were developed, and are just shipped with such software.

In MBR era, a list of one-byte numbers was used to represent partition types. But the growing number of different filesystems and their purposes exposed the outrageous non-scalability of such a solution. So GPT developers decided to abandon the problem for near and far future, and replaced them not with two-bytes or even four-bytes codes, but with GUIDs right away. They still are a fixed list, but they are much more sparse and not ordered.

2 members found this post helpful.

So it seems that I mixed up unique partitions GUIDs with partition types GUIDs. Right?

Universally Unique identifiers are usually called UUID's.
At least the blkid command lists the UUID's of my disk partitions.

GUID's sounds very similar to GID's.

1 members found this post helpful.

Thanks for the "blkid" command hint. It looks similar to "lsblk".
However, following bodqhrohro's remark, I don't understand now the difference between "lsblk -o parttype" and the PARTUUID info from "blkid -o full". The sequences are not the same so there must be a difference somewhere...

PS: the difference betwwen GUID and UUID seems very tenuous. Some sources indicate that GUID would be more Windows specific.

Not really unique in theory. In practice yes.


https://www.quora.com/What-is-the-di...-UUID-and-GUID seems to be a rather good explanation.


How to use them. https://betterexplained.com/articles...uide-to-guids/

1 members found this post helpful.

Yes.

Technically they are almost the same. You won't see any difference if not checking them with a very strict pattern. And even the name of GPT stands for "GUID Partition Table", so I don't see any reasons for doubt.

1 members found this post helpful.

Thanks for all your answers.

Then can one says that:
returns the partition type GUIDs/UUIDs (so not unique and even referenced publicly)
whereas
returns the partition GUIDs/UUIDs (a priori unique)
?

It seems that GUIDs/UUIDs are quite interchangeable. Even the RFC 4122 states:
Finally, can one states that what define a GUID/UUID are in the end the following?

• the goal of having a pretty unique big sequence,
• the format (groups of 4-2-2-2-6 bytes),
• some conventions.

Indeed, https://betterexplained.com/articles...uide-to-guids/ suggests that the implementation could be very open/free...
So am I totally wrong if I pick a MD5 (let's say d6aa97d33d459ea3670056e737c99a3d), add some dashes to it (d6aa97d3-3d45-9ea3-6700-56e737c99a3d) and then call it a GUID/UUID?

How did you get this md5 issue?

I'm not having any MD5 issue, I'm wondering what GUIDs/UUIDs represent on my computer (lsblk / blkid) and how to get them and deal with them

A UUID is an almost certainly unique number.
They can be used to index hardware resources without hitting a duplicates problem.

1 members found this post helpful.

Indeed, thanks but please see post #8 for more specific questions I have

I think its just a few formatting hyphens to aid humans.

I'm sure that a "real" UUID has no hyphens, since it is only used for comparison purposes and comparing two big numbers is much easier than comparing two strings with the hyphens in.
Personally, if I ever need to check a UUID is correct (in a fstab file) I just check the last 3 or 4 characters match with the disk partitions UUID.

I never generate UUID's - but you can do this, with the uuidgen command (which is part of libuuid from the e2fsprogs package).
See https://www.tutorialspoint.com/unix_...ds/uuidgen.htm

on my system there's 3 commands:
you (op) should look at their man pages and play with them.

Interesting thread ... Made me look up the Wikipedia entry someone mentioned above ... Pretty well written and informative - I recommend reading it.

Cheers.