Don't follow this if you later want to add some FlashRom to your expansion.
Absolutely! But ... if all that someone is interested in is a simple RAM expansion, then I believe that the suggestion is a good one since it might make routing the traces easier.
Heck, with SzörG's lovely SDcard interface and enough RAM, then modern users might not even see much need to ever add some FlashROM.
It all depends upon what each individual owner wants to do with their Enterprise.
I am trying to search the gerbers of Saint's Ram expansion(asking him...). Or the ones of the Spanish clone(I still have not contacted "Quest", the guy that copied it). If not found, don't you think it is easier to not reinvent the wheel? the layout seems very easy to design on Eagle or other similar program.
I have absolutely no interest in "reinventing the wheel", nor do I believe that I have the knowledge, experience or skills to somehow invent the "best" internal RAM expansion.
I just want/need an internal RAM expansion for my Enterprise 64, and none seems to be currently available anywhere for me to buy, so I am trying to come up with an *easy* solution that requires minimal soldering or programming, that might be easy and cheap to order, and that hardware-novices could re-order themselves in the future.
When it comes to Saint's lovely design ... the GAL that he used has been obsolete for some time, and they mainly seem to available as "used" units from China. I can find some sellers of NOS parts, but they are 20x the cost of a surface-mount 74HCT138 from JLCPCB.
Other 5V GALs, such as Atmel's ATF16V8B, are still widely available and affordable, but then that would be a change in Saint's design, and I don't know if that would work.
Then, whichever GAL is used, someone has to have the right equipment, and then take the time, to program it.
You, Saint, Kotek and others have far more hardware knowledge than I will ever have, and if you, or someone else produces some boards (of whatever type), then I will probably buy one.
But if not, then is there some reason to avoid the simple 2-chip solution that I am suggesting?
With my limited knowledge, it seems to be almost identical to what both @dangerman and @Zozosoft have already had working, and have shown in other threads.