The problem with the IBM1620 is that it has extra bits in the characters that can't be used to store data. This mean that implementing the "C" language in way which allows it to perform meaningful work possibly on existing data would be challenging. I can't see any point in implementing any language just to show it can be done... Dave G4UGM -----Original Message----- From: cctalk-bounces at classiccmp.org [mailto:cctalk-bounces at classiccmp.org] On Behalf Of Eric Smith Sent: 30 September 2014 08:51 To: General Discussion: On-Topic and Off-Topic Posts Subject: Re: Who is the world's oldest working programmer? On Mon, Sep 29, 2014 at 7:06 PM, Sean Conner <spc at conman.org> wrote: Actually, if you *can* address a single BCD digit, that will make C hard to support, because C requires that all data types have sizes of a multiple of the size of the character type, and the character type is required to have a range of at least -127 to +127 (for signed char) or 0 to 255 for unsigned char. It would seem that the easiest way to do that on a BCD machine would be to use groups of three digits (or more) as a char, and have a lot of the possible machine values of those groups be illegal. This is essentially the same reason that a compliant C on a PDP-10 would have to use 9, 12, 18, or 36 bit characters, and not directly support the normal PDP-10 native 6-bit and 7-bit characters.