Another comment for Clause 11 -- reproducibility

[Michel Hack (1-914-784-7648) <hack@xxxxxxxxxxxxxx>]

Technical  60  11 34

    Another source of variability is the max precision H supported
    for conversions between binary formats and decimal character
    strings (5.12.2).  The "H should be unbounded" should be
    honoured by making it a "shall" in reproducible mode.  That is
    because the alternative -- guaranteeing that string->binary
    conversions will never be presented with a string of more than
    M+3 significant digits -- may in fact be more difficult to
    enforce, or even to expect the programmer to observe.  (It is
    probably ok to expect programmers of reproducible-mode programs
    not to request conversion of more than M+3 digits on output, i.e.
    in the binary->string conversions -- but 5.12.2 does not provide
    for asymmetric bounds.)

    Add bullet:
    - conversions to and from external decimal character sequences
      shall not limit the maximum supported precision H (see 5.12.2).

Michel.

P.S.  I have submitted my other comments, taking the discussions here
      into account.  I'll wait a few days before I submit this one,
      just in case somebody shoots a major hole into it.
Sent: 2007-10-18 19:49:29 UTC