The Jargon File (version 4.4.7, 29 Dec 2003):

MFTL
 /M?F?T?L/

    [abbreviation: ?My Favorite Toy Language?]

    1. adj. Describes a talk on a programming language design that is heavy on
    the syntax (with lots of BNF), sometimes even talks about semantics (e.g.,
    type systems), but rarely, if ever, has any content (see content-free).
    More broadly applied to talks ? even when the topic is not a programming
    language ? in which the subject matter is gone into in unnecessary and
    meticulous detail at the sacrifice of any conceptual content. ?Well, it was
    a typical MFTL talk?.

    2. n. Describes a language about which the developers are passionate (often
    to the point of proselytic zeal) but no one else cares about. Applied to
    the language by those outside the originating group. ?He cornered me about
    type resolution in his MFTL.?

    The first great goal in the mind of the designer of an MFTL is usually to
    write a compiler for it, then bootstrap the design away from contamination
    by lesser languages by writing a compiler for it in itself. Thus, the
    standard put-down question at an MFTL talk is ?Has it been used for
    anything besides its own compiler?? On the other hand, a (compiled)
    language that cannot even be used to write its own compiler is beneath
    contempt. (The qualification has become necessary because of the increasing
    popularity of interpreted languages like Perl and Python.) See 
    break-even point. (On a related note, Doug McIlroy once proposed a test of
    the generality and utility of a language and the operating system under
    which it is compiled: ?Is the output of a FORTRAN program acceptable as
    input to the FORTRAN compiler?? In other words, can you write programs that
    write programs? (See toolsmith.) Alarming numbers of (language, OS) pairs
    fail this test, particularly when the language is FORTRAN; aficionados are
    quick to point out that Unix (even using FORTRAN) passes it handily. That
    the test could ever be failed is only surprising to those who have had the
    good fortune to have worked only under modern systems which lack
    OS-supported and -imposed ?file types?.)


The Free On-line Dictionary of Computing (30 December 2018):

My Favourite Toy Language
MFTL

    (MFTL) Describes a talk on a programming
   language design that is heavy on syntax (with lots of
   BNF), sometimes even talks about semantics (e.g. type
   systems), but rarely, if ever, has any content (see
   content-free).  More broadly applied to talks - even when
   the topic is not a programming language --- in which the
   subject matter is gone into in unnecessary and meticulous
   detail at the sacrifice of any conceptual content.  "Well, it
   was a typical MFTL talk".

   2. A language about which the developers are passionate (often
   to the point of prosyletic zeal) but no one else cares about.
   Applied to the language by those outside the originating
   group.  "He cornered me about type resolution in his MFTL."

   The first great goal in the mind of the designer of an MFTL is
   usually to write a compiler for it, then bootstrap the design
   away from contamination by lesser languages by writing a
   compiler for it in itself.  Thus, the standard put-down
   question at an MFTL talk is "Has it been used for anything
   besides its own compiler?".  On the other hand, a language
   that *cannot* be used to write its own compiler is beneath
   contempt.

   Doug McIlroy once proposed a test of the generality and
   utility of a language and the operating system under which
   it is compiled: "Is the output of a Fortran program
   acceptable as input to the Fortran compiler?"  In other words,
   can you write programs that write programs?  Alarming numbers
   of (language, OS) pairs fail this test, particularly when the
   language is Fortran.  Aficionados are quick to point out that
   Unix (even using Fortran) passes it handily.  That the test
   could ever be failed is only surprising to those who have had
   the good fortune to have worked only under modern systems
   which lack OS-supported and -imposed "file types".

   See break-even point, toolsmith.

   (1995-03-07)