In fact there is a ROWID for each row, but it's usage has been disabled in V2R5, mainly due to Oracle users abusing it :-)
The keyword ROWID was implemented in V2R3 and was supposed to be used for infomational purposes.
But Oracle users starting using it for accessing data (because it's the physical location on disk in Oracle). Despite telling them that it's a logical ROWID in TD (and such not faster than a PI-access) and making access a slow Full Table Scan they still kept using it.
In V2R5 the size of the ROWID changed from 8 to 10 bytes and those users were complaining, that their application broke.
So it was switched off in one of the early V2R5 patch levels (it can still be switched on, but this is usually not done).
Currently the keyword ROWID is only allowed within the definition of a Join Index.