Re: dynamic memory allocation

[mfb@mbunix.mitre.org (Michael F Brenner)]

To: Stephen.Leake@gsfc.nasa.gov
Subject: Re: dynamic memory allocation

Some discussion on very conservative memory allocation (either no
dynamic memory allocation or allocation only at the beginning
and then no de-allocation) can be seen at:

   The OberonT Language
   Hard Real-Time Systems
   Cache Coherency

Unfortunately, there is very little reasearch done (outside of NASA,
MITRE, and the rest of the hard-real-time community) to rigorously
prove ways to make something last many years. They just dont need it.
Consumers accept text editors, operating systems, Net browsers,
spread sheets, command and control systems, e-mail software,
airplane avionics, fragmented disks, traffic lights, billing systems,
government services, databases, year 2000 problems, and teller
machines that Hang the Mouse (that is, run off the end of a
hanging pointer or sieze due to memory fragmentation). For example,
most managers of messaging systems suffering from Year 2000 overflows have
no intention of changing the messages from 2-digit
to 4-digit years, because they feel they can afford the consequences.

Mike mikeb@mitre.org