The C language uses the dot to access members of struct and union. These members must have an identifier and identifiers are not allowed to start with a digit.

You could however do something like this while staying within the language restrcitions:

typedef union
{
  struct
  {
    int b0:1;   //use bitfields to access the bits
    int b1:1;
    int b2:1;
    int b3:1;
    int b4:1;
    int b5:1;
    int b6:1;
    int b7:1;
  } bits;       //in C99 it's allowed to leave this identifier out
  unsigned char byte;
} SfrByte;

SfrByte P0;

P0.byte = 0;    //access the whole byte
P0.bits.b0 = 0; //access bit0
P0.b1 = 0;      //access bit1 in C99 anonymous struct

This is kind of the approach IAR took in their newer versions of their compiler.

There is nothing in the way of extending the C language to support struct member identifiers containing a starting digit. But it is an extension. Then again so are the keywords 'data', 'xdata', 'code', etc. And the solution Keil chose to access bits does violate the C language IMHO:

sbit P0_1 = P0^1; //to me this means P0_1 is initialized with P0 XOR 1

Maarten