Doubtful about kerning calculation

[Morula Type]

I've used KernOn for some years now and I still get a bit doubtful about some suggestions the tool gives me but more importantly about the inability to override them if I deem them incorrect. This is a prime example: correctly (in my opinion) setting LH and VV to 0 results in +100 between K and T.
It's ok if the program doesn't agree with me, but I'd really like an option to tell him "hey, just trust me on this one", and red checks sometimes prevent that or worse, erase dozens of good pairs because of a single disagreement.

[Morula Type]

Another example: I've set AA, FH and vv to 0. VV getting +132 is very hard to accept as logical.

[Tim Ahrens]

Sorry about the late replay, you caught me while I was on holiday.

Some thoughts on this:

• Model suggestions are just suggestions. Skip those you don’t consider useful. This is how it is meant to be used. I typically end up setting about half of the suggestions as models and leave the others as auto.
• What we are seeing in the screenshots are warnings, and a removal notification. The warnings are just warnings. If you are really sure about your models then ignore them.
• Regarding the VV, it is difficult to see whether it is really to tight (or LH too lose) without seeing them in various combinations. This is something Kern On does not do for you (yet), you need to type in some test words, or other combinations, so as to visually judge whether the spacing is even. As the designer, you know your font, and you can surely think of relevant combinations. For the RSB of the L, you can simply try LHLHLHLHLHLH and see whether LH visually forms groups, or HL forms groups, or whether it is even. For VV, I would use something like OIVVE or LIVVLI or TIVVLE or similar.
• In the case of LH vs KT, sorry, I completely agree with Kern On that KT is far too tight, not only compared to LH but also as such. If your visual judgement is different then there is not much we can do about this.
• It seems that your reasoning is, “This positive autokerning is far too loose. In order to fix it, I need to set it to zero.” Why does it have to become zero? Maybe it does not have to be zero, but just a smaller positive value? In a case like this, with overly loose positive kerning, I just use the slider and drag it to the left, which turns it into a model. Usually you can do this several times as it gives you some more leeway (which is intentional). This way, you can quickly find out how tight Kern On will accept without contradicting something else. Also, the pair shown below the slider is the one that determines the minimum value without contradiction.
• For “hey, just trust me on this one", use independent pairs.
• Regarding the bottom example, the question is not whether +132 for VV is “logical” as such. The question is whether we need this positive kerning so that it is not too tight, mostly compared to the FH. This boils down to Thinking vs Looking. To me, the VV with +132 certainly does not Look too loose, at least compared to the FH. Kern On only cares about consistency, nothing else. If you Think that VV should not have positive kerning (for which there is no rational basis) then you need to fix something else. It’s difficult to tell from the pairs in the screenshot but probably the RSB of the F is too loose. I’d check this first, using FHFHFHFH or similar (see above). Then, if Kern On still insists on significant positive kerning for VV, maybe the sidebearings of V and A are just too tight? That’s my suspicion in this case. Try combinations such as VIH OIV VIO FIV TIV AIO EIV, where the I should always be visually centred. Sometimes the sidebearings are wrong, and Kern On just reveals this problem. After fixing the sidebearings you may be able to get your “logical” desired zero-pair for VV.