Where should the type information be 138

It wasn't that simple. The problem was that the status quo was badly places. In particular, many of the "PC" compilers (by which I don't mean just MS-DOS ones) were very different from UNIX "K&R" C, and the operator precedence was one aspect in which some were.

It pays lip service to it, but C99 breaks clean, portable, conforming (and even strictly conforming) C90 code in some moderately serious ways, and the debates on the reflector indicated that a significant part of the reason for people wanting to do this was to enable some implementations to avoid having to change their extensions.

This was one of the main reasons that the UK voted NO.

It wasn't, as Dennis Ritchie has implied. None of the Quiet Changes in C90 were decided without good reason, and this counts as one. If I recall, there were at least three distinct parsing models (in this area, and many others elsewhere), and the standard had to choose ONE.

There were major defects in the C90 standardisation process, but this was not one of them. Whether it was a technical mistake is a lot less clear, even today. I remember Algol 68 having the same problem ....

I was active in the BSI C Panel for the whole period of C90 from the first Public Comment until after standardisation. Round about 1997, I purged all of my documents on the matter, which I regretted later, or else I could provide documentary evidence of the situation being precisely as Dennis Ritchie said.

Regards, Nick Maclaren.