For your problem, I strongly suspect that:
\ifx\pdfpagewidth\undefined
\else
\pdfpagewidth 8.5 true in
\pdfpageheight 11 true in
\fi
is, for all intents and purposes, quite fine enough. No need for complications.
Solutions based on a \csname..\endcsname approach have the drawback that they actually create a control sequence with meaning \relax, if the control sequence did not exist beforehand. Hence people do tricks with a bunch of \expandafter's to have this meaning exist only briefly and then disappear after the \ifx test. However the TeX memory will have a new entry, even for such a briefly existing control sequence, as also does the simple fact of writing \pdfpagewidth in your source code (if not commented out).
Unfortunately as soon as some code does the \csname..\endcsname approach without the alluded-to precautions, the simple \ifx\pdfpagewidth\undefined test above will become dysfunctional. And this is the main reason, I guess, that people are reluctant to recommend it. One should be aware that LaTeX's \@ifundefined creates precisely this problem. But on modern installations, latex executable is using pdftex binary (with in particular e-TeX extensions enabled), hence this is not Knuth TeX. On more ancient installations naturally you can have Knuth TeX + LaTeX format.
Note also that sometimes one may think (depending possibly on how your IDE presents things) that one is executing the Knuth TeX tex, whereas in fact it is pdftex in dvi output mode.
The \ifx\pdfpagewidth\undefined could also be compromised by some earlier code which has given a meaning to \undefined, but that would be very condemnable thing.
\ifx\pdfpagewidth\undefined \else ... \fi. This doesn't rely one-TeXextensions. It will work also withtexexecutable. – Feb 10 '16 at 07:42