topic Re: UUID - multiple issues in Operating System - OpenVMS

UUID - multiple issues

Derek Garson — Wed, 22 Jul 2009 04:11:49 GMT

With regards to SYS$CREATE_UID there appear to be multiple issues. (The reader is assumed to be familiar with RFC4122.)

1. When getting the MAC address it uses the overridden MAC address, not the default MAC address. Hence if the chosen ethernet adapter is running DECnet, a significant assumption of version 1 UUIDs is violated.

2. The byte order in the three time fields seems to be wrong. The spec says high-low (which is the opposite of platforms that VMS is used to but normal on the internet). It seems to me that the fields are returned low-high.

3. It isn't using the clock sequence field as intended. This isn't a problem as such.

4. The service is lacking functionality e.g. can't generate version 3, 4 or 5 UUIDs (version 1 UUIDs are considered a bit dubious for privacy and security reasons) and e.g. can't return the UUID as a string. A new optional second argument, being an itemlist, would do nicely.

5. f$unique(), which seemingly just invokes this service, might have been better to return its result with bytes reversed, at least as an option.

As it is, the MAC address is always the first 12 bytes of the result and hence, when used to create file names, you don't get a good spread of names.

If functionality were added to the service, as per item 4, then some or all of that should then be reflected in new functionality in f$unique().

Re: UUID - multiple issues

Ian Miller. — Wed, 22 Jul 2009 09:06:57 GMT

Interesting.

Have you reported this to HP?

Re: UUID - multiple issues

Richard J Maher — Wed, 22 Jul 2009 10:44:58 GMT

Hi Derek,

Alan Potter was aware of similar issues about 15 years ago (including one I believe that would crash the system under load?)

I personally lodged an issue with $getdti where enteriung a wildcard lookup would only result in the wildcard portion of the response :-(

Perhaps not as important as IPsec but certainly less than 64bit addressing for RTR X86.

Good-luck!

Richard Maher

Re: UUID - multiple issues

Derek Garson — Mon, 27 Jul 2009 02:59:46 GMT

>Have you reported this to HP?

No, I know of no way of doing that myself.

Is anyone able to confirm either item 1 or item 2? (with the understanding that in order to confirm item 2 you would need familiarity with the cited RFC)

In case anyone does attempt to confirm, I am running V8.3-1H1 on Itanium hardware.

Re: UUID - multiple issues

Steven Schweda — Mon, 27 Jul 2009 04:07:12 GMT

> No, I know of no way of doing that myself.

I make no claims for its effectiveness, but:

http://h71000.www7.hp.com/

"Send us your comments":

http://h71000.www7.hp.com/fb_business.html