Manifest model issues

[ibaldin]

There are several problems in the way that manifest RDF models are constructed that make it difficult to perform visualizations and RSpec conversions.

• Manifest frequently relies on physical interface names, rather than creating virtual interface names that point to physical ones.

  • One of the way the problem shows itself is RENCI-NRIG/exogeni#163 - Flukes isn't able to find a unique path between domains because interface names are reused across different nodes. Note that the underlying provisioning problem has been fixed, however the visualization problem remains
  • Another problem is VLAN interfaces names are not sufficiently unique in some circumstances #160 with similar implications

• A separate problem is the way in which paths for stitchports are constructed, that sometimes have node-link-node-link and sometimes node-node-link-node-link sequences, as described in Inter-rack slice with stitching port does not show manifest correctly in Flukes #133

• Node naming in nodegroups Node (re)naming when generalized node groups are used  #6

After making these fixes it will be necessary to adjust Flukes, AHAB and GENI RSpec manifest converter to make sure they produce correct results.

Also it is necessary as part of fixing this ticket to document the established conventions in https://github.com/RENCI-NRIG/orca5/wiki/Manifest-Models wiki

We should also do #136 at the same time.

[ibaldin]

I would reconsider using sameAs relationship currently used as a mechanism to point from child to parent interfaces. Since sameAs has some standard OWL entailments, turning on inference on a manifest will have the undesired result of inserting sameAs relationships pointed from parent to child (needs to be verified?) due to its symmetry.