Hi,

In message "Re: Roundoff problem with Float and Marshal"
    on 03/04/18, Rudi Cilibrasi <cilibrar / ugcs.caltech.edu> writes:

|nobu.nokada / softhome.net wrote in message news:<200304171653.h3HGrgHQ002407 / sharui.nakada.kanuma.tochigi.jp>...
|> Since it was wrong a little, I fixed and committed it.
|
|Hmmm interesting numerical developments.  I continue to test it by
|doing
|a cvs update (with and without the DBL_DIG patch) and do see perhaps
|some further problems:

<snip>

|produces many PROBLEM: lines on my system.

It's the limitation of base 10 representation of float values that
Marshal uses.  We need to use binary representation like XDR to solve
this.  I'm not sure if it is worth for the cost of its implementation
and portability.  Float tend to be inexact, after all.

							matz.