Among 3 of out of the final 6 missing fibers, I think, are "empty" when the slit is connected to PFI, but not empty when it is other light source (e.g. the all fiber lamp). How should the fiber is defined?

The other 3 should be empty (the missing fibers in FCA4). I'll modify it.

Right: they could get light from anything other than Cable B. I guess I would define them up to the mtp(A) connector. Then the light sources which plug into that connector (DCB, all-fiber) would define which fibers at mtp(A) they illuminate. Does that sound right?

I thought that by design, both slit 3 and slit 4 only see 597 illuminated fibers from Cable B (per the description of D0-1,2,3-3 and 3-4 on page 18 of the latest LNA document). And that is what triggers the six missing fiberHoleIds. No?

Ah, OK I think this is the case ... the fiber map of DCB is linked to fiber hole ("fh") or MTP(A) independently  (arnaud.lefur , could you confirm?) . So I think it OK to just define them as "emp", too.

Your thought is correct, except for the slit number. A while ago, when we found FCA4 misses 3 fibers, we re-considered fiber connection. The new connection was discussed around when the Cable B1 installed on the telescope. So the slit 1 and 2 see 597 illuminated fibers from the Cable B, in the current design. (I updated it with INSTRM-1195).
(LNA's document doesn't define fiber connection among A, B and C explicitly, so I didn't ask them to update the document. )

That's correct Craig, that the missing fiberHolesId. and if my understanding is correct cable B has also all the fibers, only cableC does not have them so they should be propagate until mtpC I think.

Yuki,

I think Graham use MTP for his mapping and consider SM1 so it is equivalent to fiber hole. see

https://sumire.pbworks.com/w/file/121359156/Dummy%20Cable%20B%20Summary-GJMurray-20161121.pdf but don"t use that because there was a changed after that.

LNA mapping use the spectrograph number to generate it, so technically it is not correct anymore as there was the change of cable B connection, no ?

I will suggest that we do edit the LNA doc, to at least remove incorrect facts. Certainly the carefully curated grandfibermap should be definitive; adding statements pointing to the GFM might be enough.

I'm sorry, I was wrong. Since ID of each fiber uses PFI  information (pid), we should ask LNA to update their document. 
(Some cobra modules share two spectrograph, so mtp(A) changes for a given spectrograph)
In that case, I'll also ask them to add connection diagram among the cable A, B and C.

1. Note that I'm trying to revise GFM document with Jim, covering the recent modification. 

fmadec
Yes, Graham also has his mapping, which was suggested by Sandrine when he repaired DCB1:
https://sumire.pbworks.com/w/file/126185363/DummyCableB-SlitMappingOptions-20180501-2.pdf

What I thought is when you make pfsDesign for DCB you don't use grand fiber map... as dcbActor or another has information...

Anyway, I'll add 6 missing fibers as "empty" fibers.

All changes look fine to me.

Merged to master.