Hi All,
I happened to face the trade off of whether or not using 8255. The obvoius reason was the need 4 more I/O pins. Derivatives with more than 4 ports are not reliably available here. Thanx to Jan Waclawek and Anders Sandström for their nice sujjestions on Placing the PPI in the uppermost part of addr. space and using 8-input NAND gates in conjunction with NOR gates to construct a nice 'n simple decoder circuit with virtually no glitches and little propagation delay. The best answer from my point of view -i.e not having access to new derivatives and not being in the right mood for CPLDs- is an 8255 put on addresses FFFCH through FFFFH with the addr. decoder compromised of a couple of 8-input NAND gates found in 74HC30 plus a couple of 2-input NOR gates in the next stage from which the /CS of FLASH and PPI is provided. I went through the threads on not using 8255. It seems that multi-threading is not just an operating systems' concept :). Either I did not pay enough attention (i.e. I scrolled the texts carelessly) or (more seemingly) I am stupid enough not to get the point "like others". The main reasons of not using 8255 and my answers follows:

1) That part was never designed for use with the 8051.
- Well it unexpectedly happenned to match it quite usefully.

2) They were actually for the 8086 -so not even designed for an 8-bit processor!
- As far as mode 0 is concerned, 8255 is simply great for 8 bit operations.

3) made to match the x86 uCs and a lousy match for the '51
- The fact that it was made for x86 is not a reasonable argument alone.

4) It is obsolete, and it has been obsolete for very many years now!
- When no derivative with more I/O pins is available, PPI is still the best choice.

5) 8255 was developed about three decades ago
(when the king of diamonds were still a jack)
- Wheel was invented some 3 millennia ago.

6) Using IE is akin to using 8255's in new designs.
- If only I could download the PLCC68 version of AT89C51RD2, just like what I did with Mozilla FireFox to replace InternetExplorer

Using HC573s instead of PPI is very classy indeed and does give the circuit a professional view, but taking into consideration the extra level of decode needed for say three HC573s (I mean the need for including A1 in addition to A2..A15 in the decoder aassuming that we want the PPI or the equivalent set of latches placed in the top 3 bytes of memory without any "alias"s) makes it a needed-to-be-carefully-thought choice.

8255 might not be the best friend of 8051's great grand children, but let's not make it a taboo word.

Best.