I/O ports are quite messy... (Detect OpenMSX)

But what I find also strange is specifications of the S-3527 mentions only VDP ports 98H & 99H
Some machines like NMS-8280 and 8250 have V9938 and S-3527..
So you cannot depend on the specifications of the MSX-ENGINE.

I think the entire 98H-9FH block is reserved for the VDP, the unused ports are mirrored. So that the chip select signals provided by the S3527 can be used even if there is a newer VDP which uses additional I/O ports.

So for a machine with a S3527 and V99x8 (e.g. Philips NMS 8250), 98H-9BH will be mirrored once to 9CH-9FH. Whereas a machine with S3527 and TMS9918 (e.g. Philips VG-8020/19 or Yamaha SX-100), 98H-99H will be mirrored three more times to 9AH-9FH.

But it does? out (c),r is actually out (bc),r. Just the mnemonic is misleading, but as I said, there are several Z80 based systems which do make use of this (documented) 16-bit I/O address space, and even on MSX it is a documented possibility by ASCII.

Quote MSX Datapack:

Afaik S3527 was a part made available before MSX2 came around.

Not by definition, but in the case of machines using the S3527. Since it does not consider address bits 0-2, it leaves decoding of A0-A2 up to the VDP that is wired up. If the S3527 had decoded A1-A2, only 98H and 99H would be accessible and the chip would be unusable for MSX2 machines with V9938.

Which is why I say, ASCII probably specified the entire 98H-9FH range as reserved for the VDP, even a possible V9978 could use 9CH-9FH if it needed to and still use the S3527.

So for a generic MSX-SYSTEM / ENGINE chip this makes sense, to not decode those bits and allow to mirror. For machines with their own decoding logic, they might as well check specifically just for the I/O ports used by their specific VDP, since it is better not to mirror unnecessarily.