Forum OpenACS Q&A: speaking of driver patches ...

Collapse
Posted by Don Baccus on
Henry Minsky did something I'd been wanting to do for awhile but never
got around to doing - change the PG driver to stream BLOB content
directly to the user connection rather than first spool it to a file.

The spooling code previously in the driver came fairly directly from
the then-current Oracle driver from aD.  They later removed the
spooling and Henry's patch essentially lifts it, massages it, and
drops it into the PG driver.

Why get rid of the spooling?  Because the call that returns the file
to the user's connection isn't asynchronous anyway.

The sequence was copy blob to file, upload file to user, wait wait
wait, return to caller (the Tcl script that called the driver).

Now it's upload file to user, wait wait wait, return to caller.

I assume whoever first wrote the spooling code in the Oracle driver
wrongly assumed that control would be returned to the driver
immediately rather than have it wait until the upload was complete.

Thanks to Henry for supplying the patch, which I've now committed.