BLUF: After further consideration, I am willing to go along with a prevailing 
consensus for allowing blank nodes in SPDX 3.0  under the specific 
skolemization conditions agreed to on the call if such a prevailing consensus 
exists.


All,

I put some more thought into the blank nodes situation after our call yesterday.
I still believe we should avoid the use of blank nodes given their problematic 
nature and added complexity.
That being said, given what I believe was our consensus agreement on the tech 
call yesterday that IF we decide to allow blank nodes it would ONLY be allowed 
at the serialization level and skolemization (converting localized blank node 
IDs into globally unique IRI IDs) MUST be performed as part of ANY 
deserialization then I am willing to go along with a prevailing consensus for 
allowing blank nodes under these conditions in SPDX 3.0 if such a prevailing 
consensus exists.

This is in the interests our common goal of moving 3.0 along as quickly as we 
can and is made possible by the fact that required Skolemization removes 
showstopper integrity issues and leaves only more tractable complexity, 
performance, and practicality issues.

I still see only one advantage of using blank nodes which is brevity of 
serialized content, however, given the community’s current sensitivity on this 
issue it seems like its priority should likely be higher than some of the other 
limiting issues in play here. I suspect that as adoption and use of SPDX 
matures and adopters recognize and seek out more mature and advanced 
capabilities to utilize SPDX data the importance of brevity will likely decline 
in favor of more practical and capable approaches. We can let the community 
push that maturation forward after 3.0 release.

I do feel it important to note that deciding to allow blank nodes in 3.0 does 
bring with it several remaining concerns including increased complexity and 
resource cost for SDPX content consumers, significant limitations on linked 
data use cases, limitations on the viability and value of utilizing 
deterministic global IDs to auto-correlate content, and increased complexity in 
serialization/deserialization design and implementation across different 
serialization forms. We also will still need to decide on an appropriate 
Skolemization approach.
We should keep all of these in mind and remember that there is a relatively 
simple solution (at the cost of some brevity) when the SPDX community becomes 
aware of these issues and seeks solutions.

Thank you all for the constructive conversations on this topic and your 
consideration of these concerns.

Sean Barnum
MITRE




-=-=-=-=-=-=-=-=-=-=-=-
Links: You receive all messages sent to this group.
View/Reply Online (#5277): https://lists.spdx.org/g/Spdx-tech/message/5277
Mute This Topic: https://lists.spdx.org/mt/100512930/21656
Group Owner: [email protected]
Unsubscribe: https://lists.spdx.org/g/Spdx-tech/unsub [[email protected]]
-=-=-=-=-=-=-=-=-=-=-=-


Reply via email to