Subject | Re: [IB-Architect] RE: Classic vs. superserver (long, very very long) |
---|---|
Author | Jim Starkey |
Post date | 2002-10-15T21:26:30Z |
At 04:58 PM 10/15/02 -0400, pschmidt@... wrote:
engine meta-data. The replacing the BLR generator with something
that generates execution tree should be close to a no-brainer.
Probably the biggest problem is recognition that it would be
criminal to do the switcheroo without implementing a compiled
statement cache. Happily, the runtime world was designed to
support multiple instantiates of a compiled request (runtime
objects contain offsets into an "impure" area), so most of the
effort will be bookkeeping.
Jim Starkey
>usually software,
>No insult intended, just in 20 years of working with computers, and
>I have seen too many times where trying to bend new code to work with anold data
>structure often means something doesn't work the way it should, andefficiency goes
>down the commode. What I mean by PITA factor is the obviousl PITA oftrying to
>make a new SQL only compiler work with a data structure that was designedfor
>BLR.The major change be replacing the DSQL meta-data modules with the
>
engine meta-data. The replacing the BLR generator with something
that generates execution tree should be close to a no-brainer.
Probably the biggest problem is recognition that it would be
criminal to do the switcheroo without implementing a compiled
statement cache. Happily, the runtime world was designed to
support multiple instantiates of a compiled request (runtime
objects contain offsets into an "impure" area), so most of the
effort will be bookkeeping.
Jim Starkey