At 3:39 PM 4/8/97, Benoit MARTEL wrote:
>On Tue, 8 Apr 1997, David Brownlee wrote:
>
>>
>> On Tue, 8 Apr 1997, Erich Rast wrote:
>>
>> > Thanks for your answers!
>> >
>> > I just forgot to say that I *have* 10megs of swap space on the root
>> > partition - ok, not much, but should be enough to compile 200 lines of
>> > code!!!!??? Also the gunzip problem is a bit more weird: it *sometimes*
>>
>>       Swap space and free space on the root partition are totally
>>       different things - under NetBSD swap space has to be a separate
>>       partition. If you have no swap partition and only 5mb or ram you
>>       are liable to run into unpredictable memory problems...
>>
>
>Which brings up the question: What does NetBSD do when it runs out of
>memory?
>
>Does it recover? Does it fail safely? Does it freak unpredictably?

I think that would be "freak unpredictably."  10MB of swap sounds pretty
small to me.  I thought the recommended minimum was 20MB, and gcc is a
pretty big program independent of the size of the file it's reading.

>I've always though that the inability for a simple user to bring the
>system to an unusable state without violating security was the strongest
>points of solid UN*X systems. I feel strongly enough about this to try
>and help if necessary.

I haven't been following the generic lists where fixes would be discussed,
but the heritage from 4.4BSD says this is a serious problem.

Further elucidation from the experts is solicited.

Signature failed Preliminary Design Review.
Feasibility of a new signature is currently being evaluated.
h.b.hotz@jpl.nasa.gov, or hbhotz@oxy.edu