#1 True but not really an issue, the 2 Gb limit is reached by only a few - usually graphically oriented or database - programs. And - in case for particular cases - there is 4GT tuning:
#2 False. Only a few programs actually *need* a Pagefile (and they need it only because they are programmed that way), a typical example is Photoshop.
#3 Maybe, but it is not typical, more than the actual programs it is the OS that may decide to swap pages to pagefile.
#4 Totally false. (see explanation below)
#5 Not true. Ram will be occupied by data (which of course takes exactly the same amount of space when read or written by a 32 or 64 bit OS) what it is true is that 64 bit opcodes (i.e. the assembly language instructions) tend to be double the size of the corresponding 32 bit ones, i.e. 64 bit programs are larger than 32 bit ones.But they are only a small part of a program and programs instructions occupy only a little amount of all the used RAM.
#6 It depends. It is clear that for a given printer there may be not the specific 64 bit OS drivers, but - usually - the issue is not with actually printing (as most printers do use a printing language, such as HPGL or PostScript) but rather in all the "side" or "advanced" functions (like it is often the case with all-in-ones, printer/scanner/fax and similar), it depends on the specific device and OS.
Let's talk of the page file from an historical point of view.
Once upon a time a "normal" NT (NT 4.00) system had 64 Mb of Ram. (rule of the thumb is, take the minimum needed that MS states - 16 Mb in this case - and multiply it by 4 or take the recommended - 32 Mb in this case - and double it to have a "working" system).
At the time the OS and programs often needed more than that, and the swap file was really needed, the "magic number" used to multiply the amount of real RAM to obtain the size of the pagefile was at the most 3.
With 64 Mb the machine worked fine unless you had some specific program eating lots of Ram (at the time "lots" had a different meaning)
With 128 Mb situation was better.
With 256 Mb everything was cool.
With 512 Mb the pagefile was in practice never hit.
With 1 Gb, really never.
With a machine with 32 Mb (recommended by MS) you used 3 as the multiplier for the pagefile, so the TOTAL amountof memory available would have been 32+32*3=128 Mb
As well for a 64 Mb machine you would have used 3, thus 64+3*64=256
Now think about it, you had 64 Mb Ram and a swap file 3 times that, 192 Mb.
The TOTAL amount of available memory was 64+192=256 Mb.
With a machine with 128 Mb you would have used a 2.5 factor, 128*2.5=320 Mb, TOTAL 128+320=448 Mb
With a machine with 256 Mb you would have used a 2 factor, 256*2=512 Mb, TOTAL 256+512=784 Mb
With a machine with 512 Mb you would have used a 1.5 factor, 512*1.5=784 Mb, TOTAL 512+784=1296 Mb
It is clear that once you had 256 Mb of real Ram you had the same amount as the machine with only 64 Mb of Ram and a 192 Mb swap file, if you used the "magic" number of 2x and had a 512 Mb of swap file you had 784 Mb TOTAL.
A machine with 784 Mb of RAM and no swap file would have the SAME TOTAL amount of Ram.
A 1 Gb machine with no swap file would have 30% more TOTAL memory than the maximum a rather powerful machine. (powerful at the time, with 256 Mb of RAM, i.e. 8 times what MS recommended + 512 Mb pagefile).
The pagefile starts to make no sense whatever.
Since initially the pagefile was actually needed, and it was usually much larger than RAM, the good MS guys thought that using it as a container for crashdumps was a good idea.
Since a full crush dump is a 1:1 copy of the RAM, byte by byte, the pagefile must be at least 1x the RAM (if you actually want a crashdump)+ some space for "secondary dump data" (i.e. if you want a crashdump you cannot use 1x as "magic number" but you need a little bit more than 1, 1.2 is often used for simplicity, currently the "right" number is 1x+256 Mb, see given links).
A crashdump is a totally useless feature as I can count the people capable of analyzing one on my fingers (without needing to take my shoes off ), a crash in itself is rare and usually it is (hopefully) reproducible, so you can setup a pagefile only when you want to get a full dump (though what may actually be used/useful is a minidump).
Anyway more modern Windows NT OS's have setting in the Registry to have the crash dump go to a separate file:
which is created on-the-fly only in case of a crash.
The important point here is that when the OS (and programs) will fill up to the brim all available memory (RAM+pagefile) the system will crash.
Normally this won't happen, there are warnings and automatic safeguards to avoid this, but in the case of a program "gone loose" and eating memory without releasing it, if you have a large pagefile it will simply take more time (thanks also to the page file being on a much slower device) to crash.
So, what you are doing by having a pagefile is only that of "moving" the point where the system will crash, but it will crash alright.
On the other hand, if you setup a pagefile, let's say conventionally sized 1.2x the available RAM and monitor your usage of the machine, you can have only three possible results from analyzing the results:
1) pagefile is NEVER hit
2) pagefile is sometimes hit (but for an amount much less than the 1.2*RAM you set it up)
3) pagefile is constantly or often hit (and for an amount near the 1.2*RAM size you set it up)
if #1 your pagefile is unneeded and you can remove it alright, in the case of a program that *needs* a pagefile you can reduce it to a minimum, of course you can leave it the full size but it won't give you any advantage.
if #2 your pagefile is needed but you can reduce its size (again you can leave it the full size but it won't give you any advantage).
if #3 your machine (or OS or both) is not suited for what you do on it, you should upgrade the machine (or OS and both) and have MORE RAM.
You have to understand (about your idea of "limitiing RAM and increase the pagefile size and put it in RAM):
2GB for normal RAM in Windows 7 and 2GB with a virtual disk for the swap file. Or am I missing anything?
that RAM is FAST and pagefile is SLOWER even if placed in RAM (because there is an overhead of the driver that you use to have the pagefile in RAM).
So yes, you are missing the main point.
So is the fastest variant the 2GB for normal RAM in Windows 7 and 2GB with a virtual disk for the swap file?
No, no, NO.
That will be (besides being more prone to issues/crashes/whatever) SLOWER!
The fastest (theory) might be to have the MOST accessible RAM direct and have a small pagefile on the "inaccessible" part of RAM (this will be enogh to make programs *needing* a page file to run), but in practice you won't ever notice the difference, as the pagefile will very rarely or never be hit (and again if it is hit constantly you'd better get a more powerful machine/OS and more RAM).