Thu, 15 Nov 1990 16:50:41 +0100
|
>i think it's not so dangerous as you described it. i've just checked the
>'old' vers9011 BITEARN NODES and it has most of the above mentioned
>errors too (i haven't looked into my archives when this error occured
>first).
You are right. I get my BITEARN NODES directly from my NETSERV, which
means I had no online copy of the previous versions that I could check
and, in fact, it had not occured to me that the problem might have
already existed. As I said 1.6e does not detect it, so we can well have
been running this way for a long time. Anyway I'm glad to hear we can
relax and not have to worry every hour about what might happen because of
this problem. That doesn't mean I'm happy about what happened, of course.
>of course you are right pointing to this error. i will asap update the
>VERNODES EXEC to check for this too. on the other hand we can never be
>sure, to have 100% correct data (i'm telling this for years now). in my
>opinion tools, servers etc. must be stable enough to deal with 'minor'
>errors like this one (and in fact -as described above- they have dealed
>with this error).
I don't agree. There is NO WAY I am going to spend both programming and
CPU time ensuring that everything will work fine if links are defined
several times, or only in one direction. One thing is to write code that
will not crash with slightly incorrect input, another thing is to waste
inordinate amounts of time writing code that will work normally with such
errors. You know better than anyone how much time your verification tools
take to perform sanity checks on BITEARN NODES, and still, errors slip
through undetected. It is not reasonable to expect that applications
using BITEARN NODES have to perform even more extensive checks, and that
is why I have suggested that major applications should be tested before
BITEARN NODES is released (I understand you don't have the time to do
that, this is not a complaint about your work, just wishful thinking).
Followup to NODMGT-L only.
Eric
|
|
|