As Edward Jaffe has just noted, IBM does not seem to be planning to provide support for an RMODE(SPLIT) variant that would split a program object into two segments, one AMODE(31) and one AMODE(64). This decision is, I think, a mistake. VSCR problems at the bar rather than the line are upon us, and RMODE(SPLIT64)--together with some functional equivalent of HOBSET---is an obvious, easy if non-trivial way to address them. The argument that since RMODE(SPLIT) above and below the line has been little used its analogue above and below the bar would be little used too is convenient, but like many arguments by analogy it is unpersuasive.
John Gilmore
