> I believe if you check the documentation, INSERT - SQL does
> momentarily lock the table header in order to update the record count
> contained in the header. But you are right, INSERT INTO is much superior to
> APPEND BLANK.
I reckon that indexes make the difference. Appending a blank record
writes a blank entry into all the index tags it has, then replacing the
values replaces the tag values as well, moving them down the binary tree
and probably causing some block splitting every once in a while.
Inserting a record puts its tag values right down to where they belong,
in only one move. I've measured inserting vs appending when first I saw
a .cdx index, and ratio is about 1 to 2. Never tried without indexes,
but 1 to 1.4 is my first guess.
BTW, isn't it more proper English to write "indices" instead of
"indexes"? I mean, we do say "data" and don't say "datums". I did learn
it some... wow, thirty years ago, and the professor said that English
keeps Latin plurals.