Alexander Lamb <[EMAIL PROTECTED]> wrote:
> Well, I am afraid it didn't work.
>
> Somehow, the legacy_file_format info is not "sticky".
>
The "legacy_file_format" pragma does not appear to be
sticky, but it is. The value reported back by
PRAGMA legacy_file_format
is incorrect. But the legacy file format did get set.
Mario Frasca <[EMAIL PROTECTED]> wrote:
> [EMAIL PROTECTED] wrote:
>
> >Adding DATE and TIMEINTERVAL types to SQLite would require an
> >incompatible file format change.
> >
> well, yes, that was already clear. but: where is the type of the data
> being stored? aren't there a few spare bits to use for 'future
> additions', that is, new data types? sure, a file containing date data
> would not be understood by executables where this has not been defined,
> but maybe it is possible to do it so that they see a 'text'... or maybe
> not...
>
Mario: Look back over this thread, and others before it, and
observe all the grief that gets caused by file format changes.
I've learned my lesson: No more file format changes except
too fix a serious bug.
--
D. Richard Hipp <[EMAIL PROTECTED]>
