>On 12/03/2014 21:48, Theo de Raadt wrote:
>> How many usage cases have you tested?
>>
>> All of them?
>
>Where is the regression suite for nc located?
>I don't see it anywhere under src/regress/.
>Please don't tell me you don't have it.

There is no regression suite in our tree.

Various mailing lists over the last two years will quickly provide an
impression that there is (already) substantial pain over the
divergence between 3 variants of this utility.  Convergence has not
yet occured in any form.

Some history.  The original Hobbit code was very unrefined and
non-specific, with many forks already at that time.  It was not free
enough.  One specific path was chosen out of that mess to create the
first free version in OpenBSD, hi Eric Jackson, a monkey.  Years
later, a FSF subgroup (of some sort) created a complicated version
with new features, without trying to be compliant on the base
featureset right off the bat, onlu attempting convergence later.  So
now it is a righteous mess.  If you use it, be narrow in your requests
and you get what you want.  Be more specific, and they behave
differently.

The reality is that we don't know where people are using this code,
except we know usage is exceedingly common.  The 3 different codebases
are reused, patched, and land in different places.  Certainly
OpenBSD-based nc works different when found in Debian or FreeBSD, some
people care less about unification than their specific feature.  Yet
the usage of this tool matters greatly operationally.  In spite of the
minimal cohesive featureset being so poorly defined.  Sometimes things
like this take time to resolve.  Give it another decade, unless someone
starts to care greatly.

Primarily, we do not want to make it worse for a large contingent of
people people who rely on it, at the behest of one individual's patch.
Not saying your 2nd diff is right.  But step by step you are learning
that you can't tell if it is right, either.

The standard of care should be high, but the atmosphere has not got
anyone caring enough to ensure that the standard is met.  Your diff
might make it worse.  Sorry for my harsh tones initially, but this is
a very important point.  Upon the next upgrade, someone's infrastructure
could fail on this point.

>> The standard is high.  Rise to it.
>
>Really? I couldn't tell looking at the code.
>I am working at a standard that is beyond what you can possibly imagine.

Indeed.

Welcome to the real world.

Sorry.

Reply via email to