NASA had a FORTRAN bug in 1963 that

  DO 10 I=1.10

wasn't a loop but an initialization of DO10I. The decimal point (not comma) made the spaces not matter.

In fact, nope. The Fortran compiler people found themselves in uncharted territory and thought it would be a good idea to remove all the pesky spaces before scanning and parsing, so as not to have to deal with them.

erik_seaberg's sibling comment is in connection with the situation that DOI=1.10 meant assigning the value 1.10 to variable DOI as in DOI = 1.10, whereas DOI=1,10 is the loop header DO I = 1, 10: stepping I from 1 to 10. Spaces would be written only for human readability, ignored similarly to comments. In fact, ignored harder than comments. For instance, in C, if we have DO/* comment */I, it doesn't become the token DOI; a removed comment is replaced by a space.