[BUG] graph.rank and graph.linkorder should not be copied from links to interfaces

[ipspace]

We want to have hosts attached to a network placed below the subnet box. The only solution that works today is:

links:
- r1:
  r2:
  h1:
    graph.linkorder: 200
  h2:
    graph.linkorder: 200

It would be nicer if we could change the link graph.linkorder and graph.rank:

links:
- interfaces: [ r1, r2, h1, h2 ]
  graph.linkorder: 50
  graph.rank: 2

Unfortunately, that doesn't work because graph.rank and graph.linkorder are propagated to all interfaces (so they don't influence the sorting process)

Based on #3197 reported by @a-v-popov

[a-v-popov]

If I set ranks r < h, would it make sense for linkorder to be automagically set r < bridge < h?

…

On Thu, 12 Mar 2026, 16:02 Ivan Pepelnjak, ***@***.***> wrote: *ipspace* created an issue (ipspace/netlab#3199) <#3199> We want to have hosts attached to a network placed below the subnet box. The only solution that works today is: links: - r1: r2: h1: graph.linkorder: 200 h2: graph.linkorder: 200 It would be nicer if we could change the link *graph.linkorder* and *graph.rank*: links: - interfaces: [ r1, r2, h1, h2 ] graph.linkorder: 50 graph.rank: 2 Unfortunately, that doesn't work because *graph.rank* and *graph.linkorder* are propagated to all interfaces (so they don't influence the sorting process) Based on #3197 <#3197> reported by @a-v-popov <https://github.com/a-v-popov> — Reply to this email directly, view it on GitHub <#3199>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AEDV7WOYFPIF3HL46QUHPVT4QLGPBAVCNFSM6AAAAACWPZR2XWVHI2DSMVQWIX3LMV43ASLTON2WKOZUGA3DKMRZGIYDEMA> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[ipspace]

If I set ranks r < h, would it make sense for linkorder to be automagically
set r < bridge < h?

Just because two devices have a different rank doesn't mean you want to have them on the opposite sides of the subnet box. I agree that would make sense in your scenario, but I don't want to assume that's always the case.

[a-v-popov]

You current assumption for linkorder is $r = h = 50 &lt; 100 = bridge$.