While reading through Volume 2 of the CDB 3.2 spec, I found mention of a GSModelFileSize in section A.19. I can't find where this term or value is defined. What is it?
You are right, it is missing in the text. As we are about to review the text in that section to ensure it applies to GT models as well, we will ensure the intended value for <i>GSModelFileSize</i> can apply and update the text accordingly.
As you read section A 1.19, please comment on additional improvements.
The value for GSModelFileSize in the section A 1.19 in spec 3.2 is 16MB. The missing text will be incorporated in a new edit of the spec 3.2 in the coming weeks.
The section will also highlight that the value is valid for GT models as well. Since GSModelFileSize refers to the sum of all 3D model files referred by a feature tile (as opposed to a zip file size in a GSModel tile), the approach works for both.
I would question if that is a good value to use in CDB 3.2.
Back when the draft version of CDB 3.2 was released, it used this same placeholder. From conversations with Bernie, he mentioned that 16MB was a value being discussed. But the big difference was that the draft was changing the table that mapped a model's significant size into a CDB LOD (Table 3-1, section 3.3.7 in the CDB 3.2 spec vs Table 3-20, section 188.8.131.52.1 in the CDB 3.2 draft 1). Essentially, the lods were pushed down three rows in the draft. I'm not a big fan of putting file size limits on these files (I'd rather use numbers based on the geometry, like vertex counts), but the file size limit seemed reasonable to me based on pushing the model LOD down 3 rows in the draft.
If we just scaled the value by those 3 rows, we get 1024MB (16MB * 4^3 tile size change in 3 LODs). That is too large to be practical, so what about 128MB (16MB * 2^3). I have seen lots of GS model zip files blow past 16 MB (including some that were much larger than 100MB on a badly constructed database). Also, both the Camp Pendleton and Yemen databases on the Presagis portal have GS model zip files larger than 16MB (the largest is 46.7 MB), which are admittedly small enough to download off the web and might not reflect a full CDB dataset. So for an example 46MB zip file, you could be pushing models down two levels of detail, which makes the warning on page A-62 even more applicable about not being able to trust the database's relationship between a model's significant size and its 3DModel-LOD.
Also, is this file size, as it applies to GT features, per feature tile (shapefile)? Is it a total of the geometry for each unique GT model that is used or iteratively added for each reference of the model? (i.e. if you have one building referenced 10 times, does this count once or ten times?)