Hi, team
I was told that the last 3 characteristics of ROWID stands for the row number of database block. Theoretically, every database block could store as many as 65536 rows.
I was wondering how many records could stores in one block (with the size of 8K) practically. Therefore, I did a testment this morning:

SQL> create table CH1(c char(1)) pctfree 0;

Table created.

SQL> insert into CH1 select 'c' from dual connect by level<3000;

2999 rows created.

SQL> select dbms_rowid.rowid_block_number(rowid) bn,count(*) from CH1 group by dbms_rowid.rowid_block_number(rowid);

BN COUNT(*)
---------- ----------
140574 733
140571 67
140572 733
140573 733
140575 733

SQL> create table CH5(c char(5)) pctfree 0;

Table created.

SQL> insert into CH5 select 'ccccc' from dual connect by level<3000;

2999 rows created.

SQL> select dbms_rowid.rowid_block_number(rowid) bn,count(*) from CH5 group by d
bms_rowid.rowid_block_number(rowid);

BN COUNT(*)
---------- ----------
140579 67
140582 733
140580 733
140581 733
140583 733


SQL> create table CH6(c char(6)) pctfree 0;

Table created.

SQL> insert into CH6 select 'ccccc' from dual connect by level<3000;

2999 rows created.

SQL> select dbms_rowid.rowid_block_number(rowid) bn,count(*) from CH6 group by dbms_rowid.rowid_block_number(rowid);

BN COUNT(*)
---------- ----------
140589 672
140588 672
140587 311
140591 672
140590 672

SQL>

According to my testment, it shows that only 733 records could be stored in one 8K block at most.
Here's my question:
1) Is it true that 733 records is the upper limit?
2) I wanna know where the rest space (8K - 5 byte*733 = 4527 byte) goes away?
3) Should the block HEADER(common and variable header, table directory, row directory and ITL slots) consumes all the rest space in the block? If not, could you please figure out what else factors could consume the rest space?
4) Suppose that the size of block is 32K, could the maximum of records in one block up to 65536?

block header = fixed header + variable transaction header + 
                 table directory + row directory 
 
     where: 
        fixed header = 57 bytes  (the 4 bytes at the end of the block have  
                                  already been taken account of in the 24 bytes 
                                  for the cache header) 
 
        variable transaction header = 23 * i  where i is the value of INITRANS 
                                              for the table, or can grow 
                                              dynamically up to MAXTRANS. 
 
        table directory = 4 * n    where n is the number of tables. 
                                   n = 1 for non-clustered tables. 
 
        row directory = 2 * x    where x is the number of rows in the block. 

Note: The minimum row size for a non-clustered table is 9 bytes. Therefore, if 
      the calculated value for an average row size is less than this absolute
      minimum row size, use the minimum value as the average row size.