yang-problems.txt

Overall observation: 
  RFC 6020 is a "legal" or "committee" document meant only for human consumption.  
  It is not meant to be executable (i.e. not meant to be consumed by machine to define a parser)

-----------------------------------------------------------------------------
From the InstaParse docs:

    However, in ABNF notation, angle brackets are meant to be used for prose descriptions of
    some concept that can't be mechanically specified in the grammar. For example:
        P = <a prime number>

-----------------------------------------------------------------------------
One may interepret the ABNF from RFC 6020 as a kind of "loose upper bound". That's kind of like
saying "Alan cannot jump higher than 10 feet".  This does not

  - tell you exactly how high Alan can jump
  - tell you anything about how high Alan can jump relative to other people.
  - tell you how to calculate Alan's max jump height
  - it only tells you that anything that can jump 10 feet (or higher) isn't Alan 

-----------------------------------------------------------------------------
Examples from RFC 6020:

p18, Defining "decimal64" as "64-bit decimal" is stupid. Either make it float/double, float32/float64, or BigDecimal.

p142, "This grammar assumes that the scanner replaces YANG comments with a single space character."

p162, string = < an unquoted string as returned by the scanner >

p162, 
  1. duplication: why are there 2 def's?
       identifier-arg-str  = < a string that matches the rule identifier-arg >
       identifier-arg      = identifier

      The reason is that these expressions are all interpreted the same by YANG:
           int   year  ;
           int  "year" ;
          "int"  year  ;
          "int" "year" ;

      From the RFC:  
          "The following strings are equivalent:"
          ---------------------------------
            hello
           "hello"
           'hello'
           "hel" + "lo"
           'hel' + "lo"
          ---------------------------------

  2. no (practical) way for ABNF to reject "xmlName", "XmlName", "xml-name", "XML-name", etc
   ;; An identifier MUST NOT start with (('X'|'x') ('M'|'m') ('L'|'l'))
   identifier          = (ALPHA / "_")
                         *(ALPHA / DIGIT / "_" / "-" / ".")

  3. which one is it (problem since '*' isn't greedy like in regex)
        1*("abc...xyz")
        name -> [:ident "name"]
        name -> [ [:ident "n"] [:ident "a"] [:ident "m"] [:ident "e"] ]
     
-----------------------------------------------------------------------------
Deficiencies

Yang has nothing like a typedef form C/C++. I cannot define a type "complex" { float real; float imag;}
  Grouping/Uses doesn't work
  `list` could be forced to work but the required `key` value is stupid & it's still not a typedef.
  BTW, `typedef` in yang can only narrow existing types and cannot define a record or struct (that
    is what "list" is for! Stupid name!)
  Example "brocade-rbridge.yang" shows this stupid/ambiguous example:

    list rbridge-id {
      key  rbridge-id;
      leaf rbridge-id {
        type uint32;
      }
      ... 
    }