[Date Prev][Date Next][Thread Prev][Thread Next][Author Index][Date Index][Thread Index]
New things in the backend
- To: <tribble>
- Subject: New things in the backend
- From: Ravi Pandya <ravi>
- Date: Sat, 23 Jun 90 17:08:16 PDT
- Cc: <markm>, <dean>, <michael>, <xtech>
- In-reply-to: <Eric>,12 PDT <9006230817.AA22426@xanadu>
Date: Sat, 23 Jun 90 01:17:12 PDT
From: tribble (Eric Dean Tribble)
Should they [Sensors] be dependents?
They don't need to be dependents. Partial Loaves should just destroy
asll sensors when they get destroyed. Is that the only reason they
might need to be dependents? I don't consider tracking destruction to
be part of the dependence mechanism (it's at a different semantic
level than stateful operations).
There's no problem getting rid of sensors when the PartialLoaf goes
away. I was worried about getting rid of the reference to the sensor
in the PartialLoaf when the session ends or when the Sensor is
explicitly deleted by the client program.
What happens if the loaf gets sent to disk?
The variable contining pointers to sensors should not be a copy field,
so that it doesn't get carried to disk with the loaf. I think it
should continue to exist in memory even when the loaf is on disk. It
might even make sense for it to not be a field of the PartialLoaf but
rather a separate HeaperSpace table that acts as a soft field.
This sounds like it might work, if you can figure out a way to
reassociate the list of sensor with the PartialLoaf when it comes
back. I don't think a HeaperSpace table will work because the
PartialLoaf won't necessarily have the same oop when it comes back.