Version 5.1 has added a UUID function; which returns a 36-character UUID value

presumably created from the version 5 uuid_t described in RFC 4122, with the

information (the "name") described in the PL/I documentation.

RFC 4122 indicates that a unique "name identifier" be used to create the SHA1

hash (along with the given name information) but the IBM documentation does

not describe that "name identifier".

Can that be detailed?

  - Many thanks -

  - Dave Rivers -

Hmm - replying to myself - but OK.

Also - I note in RFC 1422 regarding version 5 UUIDs, it says:

The requirements for these types of UUIDs are as follows:   o  The UUIDs generated at different times from the same name in the      same namespace MUST be equal.

I believe including the time-stamp (the STCK value? - the documentation isn't clear)

completely defeats that requirement.   If I'm understanding things correctly, a version 5

UUID is intended to be unique, but easily regenerated; so that an invocation of the UUID

builtin on the same machine (certainly multiple times within the same PL/I program)

should return the same value and not be time-dependent.  I believe by muddying

the "name" with a STCK value that behavior has been lost.

Would it be better to remove the STCK value from the "name"?

Version 1 UUIDs are time-dependent.

Correct, if you need a user supplied "namespace" as argument to uuid, it would need an enhancement. The current implementation only address the uniqueness issue with the use of time dependent value.

Well - it's not that we need a user-supplied namespace; more that using the time-value

breaks the version 5 standard that requires the same UUID be returned on the same machine.

Incorporating the current time value in the uuid-string generation breaks that.

Also - the documentation does not provide the "name identifier" used to create the UUID.

In the 5.2 LRM, we updated the description of the UUID function with this additional text

The name-space information consists of:
1. System information:
v CPU count
v MVS name
v storage size
v model and serial
2. Runtime 64 bit TOD value.
3. Job name or, if possible, the PID.
The UUID is a SHA1 hash of the above.

Hi Peter!

 Thanks for the pointer...  however, the doc says that UUID returns a version 5 UUID.

 But, doesn't including the time-of-day value in the name-space break the version 5 requirement

of returning the same UUID?   I suppose not, if the time-stamp is put into the name-space (technically,

then, each invocation of UUID is using a different name-space I suppose.)  Just curious... 

Also - let me add that a version 5 UUID is a "name" created within a "namespace"; those are 2 separate

ideas...

I'm guessing that the doc is describing the "name" that is used.

But - there is no description of the "namespace" in which that name resides.

For example, if you were going to do use the fully-qualified domain name to create a UUID, that

name is from the "DNS" namespace, which is defined as "6ba7b810-9dad-11d1-80b4-00c04fd430c8".

Is there any plan to define/describe the namespace the UUID builtin is using, so that it doesn't

possibly conflict with another namespace?

Or - is the documentation actually describing the namespace that is used, but not describing the name?

I think there are two items that go into a version 5 UUID, and I'm kinda only seeing one - maybe??