svn commit: r335402 - head/sbin/veriexecctl

[Ian Lepore]

On Thu, 2018-06-21 at 11:13 -0700, Conrad Meyer wrote:
> On Thu, Jun 21, 2018 at 9:51 AM, Stephen Kiernan <hackagadget at gmail.c
> om> wrote:
> > 
> > On Wed, Jun 20, 2018 at 10:36 PM, Eitan Adler <lists at eitanadler.com
> > > wrote:
> > > 
> > > 
> > > On 19 June 2018 at 20:08, Eitan Adler <lists at eitanadler.com>
> > > wrote:
> > > > 
> > > > On 19 June 2018 at 18:08, Stephen J. Kiernan <stevek at freebsd.or
> > > > g> wrote:
> > > > > 
> > > > > Added: head/sbin/veriexecctl/Makefile
> > > > > 
> > > > > =============================================================
> > > > > =================
> > > > > --- /dev/null   00:00:00 1970   (empty, because file is newly
> > > > > added)
> > > > > +++ head/sbin/veriexecctl/Makefile      Wed Jun 20 01:08:54
> > > > > 2018
> > > > > (r335402)
> > > > > @@ -0,0 +1,11 @@
> > > > > +# $FreeBSD$
> > > > > +
> > > > > +PROG=  veriexecctl
> > > > > +MAN=   veriexecctl.8
> > > > > +SRCS=  veriexecctl_parse.y veriexecctl_conf.l veriexecctl.c
> > > > > +
> > > > > +WARNS?=        3
> > > > Why are we introducing new code with lower-than-6 warnings ?
> > > In all the commotion about the more important issues this fell
> > > through.  Also its argument parsing appears to not be using
> > > getopt[_long] ?
> > 
> > I replied to this 2 days ago with:
> > "veriexecctl came from NetBSD originally and that is what they had,
> > but I believe it should be able to be bumped up."
> > 
> > However, there has been some discussion about just not putting in
> > veriexecctl for now and wait for some work that Simon Gerraty has
> > been
> > doing, using some of the work for the verified loader, instead.
> > However, it
> > would also mean that in the meantime, there would be nothing
> > available
> > to be able to people to try out veriexec to provide some feedback
> > until
> > that utility was completed and committed.
> Hi,
> 
> While the code is out of HEAD, it can be posted to a github branch
> (or
> a projects/ branch if you prefer SVN) for people to try.
> 
> Best regards,
> Conrad
> 

Yeah, put it on a branch where it'll get ignored for another two years.

If this code had been committed long ago, as it probably should have
been, then people would have been playing with it, and by time I needed
it a few months ago there would have been all kinds of useful info in
mailing lists and blogs about how to set it up and what was good and
bad about it and so on.  Iterative refinement would have been underway.

Instead what I found was a bunch of patches and a big steep learning
curve with no existing information about using it in the real world.
With that info available, I/we ($work) would have been in a position to
quickly adopt it and begin contributing to the ongoing refinement.
Instead I had to conclude that product deadlines just didn't allow us
to even try to get it working from a standing start as first-adopters,
so we had to move in a different direction. Even though this is a
better solution than what we did, business practicalities will likely
prevent us from circling back and changing everything over to this
scheme in the future, so now we'll end up never contributing much to
this work.

So, IMO, all this calling for things to be reverted isn't just
inappropriate, it's actively harmful. This is -current where
development happens and imperfection is expected. Hiding work in
patchsets and reviews and alternate branches and other shadowy places
because it's not perfect is just a way of ensuring it never gets any
better.

-- Ian