Summary

  • Revision 07 of a PCE working-group Internet-Draft adds explicit complete-replacement semantics for entropy-label positions in a PCUpd path update.
  • A valid update can replace or remove the old set; invalid or partial position data must not change the existing path or position state.

Consider a segment-routed MPLS path whose entropy-label pairs are inserted after two particular segments. A path computation element (PCE) revises the route. If the path computation client (PCC) retains the first old insertion point while applying the second new one, neither the former calculation nor the new calculation describes the resulting label stack. That is the transition the PCE working group's 28 September draft now tries to close.

The mechanism predates this revision. The earlier text already described capability, request and SR-ERO position flags for entropy-label insertion. Revision 07 changes the handling of a later path update: the entropy-label position information in a PCUpd is the complete replacement set for that label-switched path. A valid SR-ERO with the E flag replaces the previous positions as a unit. A valid SR-ERO without that flag removes all previously installed ELI/EL pairs.

An invalid or partially valid new set must leave both the existing path and its entropy-label state untouched; the PCC should return an error correlated with the update's SRP identifier.

The revision also tightens the negotiation boundary. A client must not request position computation from a PCE that has not advertised the capability. A PCE must not send a position set that is invalid or partial. When missing, inconsistent or insufficient stack-depth information prevents a valid calculation, the draft calls for no usable SR-ERO subobjects and an error, not a plausible-looking subset. An unsolicited position response must not be applied by the client.

There remains a drafting question worth keeping visible. A preceding sentence discusses an invalid recomputation and a path update without entropy labels, while the new PCUpd paragraph says invalid or partial position information must leave existing state unchanged. Those can be reconciled if a valid no-E-flag removal is distinguished from an invalid flagged update, but the text has not fully spelt out every ordering. Implementers should not turn that editorial ambiguity into an unrecorded choice to retain stale pairs. Proposed error values are still marked TBD.

Datatracker lists the document as an active Internet-Draft in PCE working-group last call, with I-D Exists state and intended Standards Track status. None of this makes it an RFC, proves interoperable implementations, or demonstrates improved load balancing on a deployed path.

Sources