What can be learned from a UUID?
The version and the variant — they are written into the identifier itself — and whatever the remaining bits hold. That depends on the version: versions 1 and 6 give up the creation time at 100-nanosecond resolution, a clock sequence and a node address, version 7 gives Unix time in milliseconds. Version 4 is random apart from the service bits, so there is nothing to learn from it. Versions 3 and 5 are a hash of a name, and the original name cannot be recovered from it — that is what a hash is for.
Is the identifier sent to the server?
No. Decoding runs in the browser in JavaScript: the pasted value never leaves your machine, never reaches the server logs and is not stored anywhere. For the same reason the decoder keeps working without a connection once the page is open.
Which written forms are accepted?
All the common ones: canonical groups separated by dashes, the form in braces, an address like urn:uuid: and simply 32 hexadecimal characters in a row. Case does not matter. Only the written form is checked — the decoder cannot tell you whether your own system issued this particular identifier.
Why is the node address not a MAC address?
The specification allows random bits with the multicast bit raised in place of a network card address. Most libraries do exactly that: a real MAC address would disclose the machine on which the identifier was created. The decoder marks such a node — on a real network card address the multicast bit is clear.