If whatsits are aptly named is a matter of opinion, as I think they would fit better in TeX's semantics if they were called afterallnodes; whatsits represent commands whose execution is delayed or are special commands associated with a particular device or system and are not part of TeX's normal processing flow.
It is interesting to investigate Knuth's rationale for introducing them. In a meeting with NTG members on March 13th, 1996 Knuth in reply to a question said:
I tried to make the programs so that they would have logical structure
and it would be easy to throw in new features. This hasn’t happened
anywhere near as often as I thought because people were more
interested, I think, in inter-changeability of what they do; once you
have your own program, then other people don’t have it. Still, if I
were a large publisher, and I were to get special projects— some
encyclopaedia, some new edition of the Bible, things like that — I
would certainly think that the right thing to do would be to hire a
good programmer and make a special computer system just for this
project. At least, that was my idea about the way people would do it.
It seems that hasn’t happened very much, although in Brno I met a
student who is well along on producing Acrobat format directly in TeX,
by changing the code. And the Omega system that you mentioned, that’s
150,000 lines of change files [laughter].
I built in hooks so that every time TeX outputs a page, it could come
to a whatsit node and a whatsit node could be something that was
completely different in each version of TeX. So, when the program sees
a whatsit node, it calls a special routine saying, ‘how do I typeset
this whatsit node?’ It’ll look at the sub-type and the sub-type might
be another sub-type put in as a demo or it might be a brand new
sub-type.
A whatsit can appear in either a horizontal or a vertical list and has no dimensions. It signifies an operation that should be delayed as it doesn't fit in its ordinary scheme of things. The paragraph builder and the page builder scan lists submitted to them and execute certain types of whatsit. They are useful when associated with specific implementations.
The more common whatsits are the ones associated with the main vertical list:
(a) delayed writes generated by \write. The token list of a delayed \write is not written-out until the surrounding material of a \write makes it to the output routine where a \shipout is executed. Therefore, the write token list has to be stored on the main vertical list.
(b) specials that use the \special command. The token list of a special command is stored with the main vertical list because the token list needs to be written to the dvi file. This happens, as in the case of write at the time of shipout.
Practical implementations can be found in postcript, pdf, color drivers and graphics programs. An interesting read is always the hyperref manual. The package uses \specials extensively to implement the interface between TeX commands and the PDF page description language. They are very simple to write:
\immediate\special{!pdfpagelabels #2}%
To summarize it is a free for all hook/interface. Why they were called whatsit -- my guess is that it was a Knuth (ala \fi) whatisit. This simple innocent special command enabled TeX to survive and adapt over the years, producing output from postcript to PDF and introducing color and graphics.
\whatsitprimitive. However,\writeand\specialproduce what are called "whatsit" nodes, which behave differently from all other kinds of nodes. Thinking about it, I guess\markand eTeX's\marksalso produces whatsit nodes, and perhaps\insertdoes something similar. – Bruno Le Floch Feb 02 '12 at 08:30