RE: [uddi-spec] RE: Question on use of colons in v3 key names

[John Colgrave <colgrave@uk.ibm.com>]

I meant to raise this issue at the FTF
but forgot.  I think we need to fix this for 3.0.2.  I will write
up a very brief CR for this.  Can one of the chairs give me a CR number
please?

Regards,

John
--------------------------------------------------
John Colgrave
Architect, IBM WebSphere UDDI Registry

Andrew Hately <hately@us.ibm.com>

02/09/2004 04:13

To	"Steve Capell" <steve.capell@redwahoo.com>
cc	"'Rogers, Tony'" <Tony.Rogers@ca.com>, uddi-spec@lists.oasis-open.org
Subject	RE: [uddi-spec] RE: Question on use of colons in v3 key names

John and I were discussing this issue a few weeks ago and he posted a note
to the list suggesting that we remove : from the list of characters that
are valid in KSS to remove this confusion. 

I believe this to be a necessary change in order to enforce partitions
by lexical key syntax as currently described in the specification.

The message sent by John is: http://lists.oasis-open.org/archives/uddi-spec/200408/msg00064.html

We should probably write this up as a formal change request as it seems
to have generated quite a bit of discussion today.

As for the longer term issue regarding rampant use of : characters in URNs
and portablity of the URN into the UDDI URI scheme, the keying scheme was
something we restricted heavily through the V3 change request process with
the caveat that we needed to come back and examine the use of any URI (and
thus any URI scheme) as unique keys in the future.  Since this presents
some challenges in determining uniqueness, such as which part or parts
of the URI need to be case folded for comparison, we deferred the issue
and proposed restrictions in the UDDI scheme at that time.  Do we
need to reexamine this restriction to the UDDI URI scheme and/or the corresponding
character restrictions in the context of relaxing them in v.next?


Andrew Hately
IBM Software Group, Emerging Technologies 
email: hately@us.ibm.com
phone: (512) 838-2866

"Steve Capell" <steve.capell@redwahoo.com>

09/01/2004 07:35 PM

To	"'Rogers, Tony'" <Tony.Rogers@ca.com>
cc	<uddi-spec@lists.oasis-open.org>
Subject	RE: [uddi-spec] RE: Question on use of colons in v3 key names

OK so I can see that publishing of uddi:widgetCo-com:marketing:niftyService:keyGenerator
could fail because the node will complain that the publisher does not own
the uddi:widgetCo-com:marketing
partition.  That does not stop a regular tModel publisher using colons
in a KSS provided he does not try to create another key generator that
skips a colon (if you understand what I mean).  Ie: “uddi:widgetCo-com:marketing:niftyService:version1.0”
could be a valid tModel
in the “uddi:widgetCo-com:” key domain even though trying to publish
“uddi:widgetCo-com:marketing:niftyService:keyGenerator” is invalid.  So
we COULD permit colons in KSS but not allow skipping colons in keyGenerator
publishing.   
  
One issue to note is that it will probably be quite common practice to
use existing URIs as tmodel keys.  If you look around you find very
liberal use of colons in URIs.  For example take a look at the UBL
TC’s use of namespaces for core components (which are likely to be published
to UDDI registries: 
urn:oasis:names:tc:ubl:CommonAggregateComponents:1:0-draft-10.1
  Plenty of colons there.  The UBL key generator domain would
probably be “uddi:oasis:names:tc:ubl:”
 so there are two extra colons in the KSS – not to mention the question
of whether all four of the colons in the key generator will genuinely be
implemented as higher level key generator. 
  
By the way, what is your phone number – If you don’t mind, I’d like
to run by an issue about repository meta data…. 
  
Steve Capell 
Red Wahoo Pty Ltd 
+61 410 437854

---

From: Rogers, Tony [mailto:Tony.Rogers@ca.com] 
Sent: Thursday, 2 September 2004 9:00 AM
To: Steve Capell
Cc: uddi-spec@lists.oasis-open.org
Subject: RE: [uddi-spec] RE: Question on use of colons in v3 key names
  
Two issues: 
  
You can't have the name of a keyGenerator as a tModel AND a keyGenerator
(OK, a keyGenerator is a tModel, but you know what I mean) - a given key
must be either a keyGenerator or a regular tModel, not both.
  
I'm also certain that you can't skip a level in a keyGenerator - someone
must own uddi:widgetCo-com:marketing. The publishing of uddi:widgetCo-com:marketing:niftyService:keyGenerator
will fail with an error indicating
that the publisher does not own the uddi:widgetCo-com:marketing partition.
  
I suggest raising this on the list - it's possible I've overlooked something
(but I don't believe so). 
  
Tony Rogers 
tony.rogers@ca.com

---

From: Steve Capell [mailto:steve.capell@redwahoo.com] 
Sent: Thursday, 2 September 2004 8:51
To: Rogers, Tony
Cc: uddi-spec@lists.oasis-open.org
Subject: [uddi-spec] RE: Question on use of colons in v3 key names
Tony, 
  
Thanks for your feedback. Note that KSS = “Key Specific String”.  
It seems to me that colons in the kss SHOULD be illegal from a semantic
and consistency perspective but it COULD be technically possible because
of the way key partitions work (I’m supposed to be the expert now…!?
J
 ).  The UDDI node knows which part of the string is the key
partition and which part is the kss because of the declaration of a keyGenerator
tModel.  So “IN THEORY” I could define a key generator as:
  
Uddi:widgetCo-com:keyGenerator 
  
Then create a tModel like: 
  
Uddi:widgetCo-com:marketing:niftyService 
  
And then, even though there is a colon in the ‘marketing:niftyService”
KSS, the UDDI node knows that it is in the KSS and not part of a new partition
because the original declaration of the uddi:widgetCo-com:keyGenerator
tModel.  WidgetCo could, IN THEORY, then go on to define a new key
partition as: 
  
Uddi:widgetCo-com:marketing:niftyService:keyGenerator
  
And then define new tModel is that key generator domain such as :
  
Uddi:widgetCo-com:marketing:niftyService:version1.0
  
So now we have a tModel key with three colons but only two of which represent
key partitions.  Technically feasible but pretty confusing and pretty
bad practice I’d say. 
  
So I guess what I am saying is that the permission or not of colons in
the kss is more of a policy issue than a technical issue because an implementer
could manage it without breaking the model.  Is it safe to assume
that the TC position is that colons are NOT allowed in the KSS?  
If so then we need to review all Technical notes for the definition of
v3 keys that use colons where they shouldn’t (like the ebxml TN).
  
Comments from the TC welcome… 
  
Regards, 
  
Steve Capell 
Red Wahoo Pty Ltd 
+61 410 437854

---

From: Rogers, Tony [mailto:Tony.Rogers@ca.com] 
Sent: Wednesday, 1 September 2004 7:39 PM
To: Steve Capell
Subject: RE: Question on use of colons in v3 key names
  
I'd say that you have spotted an error. 
  
I'm fairly certain that colons are not permitted in the trailing portion.
Indeed, I believe there was a comment on this very subject on the list
in August (late August, I think). The trailing portion is referred to as
KSS (I'm sorry, the definition of that escapes my tired brain at the moment)
- you might want to have a quick look for it. 
  
Tony 
-----Original Message----- 
From: Steve Capell [mailto:steve.capell@redwahoo.com] 
Sent: Wed 01-Sep-04 18:08 
To: Rogers, Tony 
Cc: 
Subject: Question on use of colons in v3 key names
Tony, 
  
I have a quick question… 
  
The colon character is used to define the partitions in a v3 key name.
  Therefore it is poor practice to use colons in the key specific
string.  However the question is – is it illegal?  
If so then there are some problems with the uddi-ebxml technical note where,
for example, the tModel representing the bpss specification is:
  
tModel Name: untmg-org:BusinessProcessSpecificationSchema:v1_10
tModel Description: UN/CEFACT - ebXML Business Process Specification
Schema 
v1.10 
tModel UDDI Key (V3): uddi:untmg.org:businessprocessspecificationschema:v1.10
Derived V1, V2 format Key: uuid:1a2a88af-54f8-316c-aaf1-e1fc2ef1c0e9
Categorization: specification 
  
Note the colon separating the version in the v3 key name.  It is unlikely
that the designer intended this colon to represent a key partition.
  
Cheers 
  
Steve Capell 
Red Wahoo Pty Ltd 
+61 410 437854

S/MIME Cryptographic Signature