>>and what about speed ?
>>I made a test and seems like append was 4 to 10 times slower
>
>I wouldn't comment on that. From test to test one might be faster than the other. My reasoning is simple:
>
>select myTable
>Append Blank
>replace field1 with m.somevalue, ...
>
>vs:
>
>insert into myTable (field1, ...) values (m.somevalue, ...)
>
>latter is a single atomic operation. Insert into also have variations like "insert into ... from memvar|array".
>
>It has been a long time I don't use 'Append Blank'.

I'd only add one more thing: since the table is very likely to have indexes, the indexes need to be updated too. With append blank, we're writing blanks into the fields and indexes, then during a replace we're doing it again (and potentially reshuffling some of the b-trees in some of the indexes). So it's not one write vs two writes, it's n writes vs 2*n writes.