[rb-general] [PATCH] Document timestamp clamping

Eric Myhre hash at exultant.us
Sun Nov 6 20:16:26 CET 2016

I have to come in out of left field here to pipe up as one of those from
the "silent"group Chris postulates.

In RB stuff I've done to date, flattening or discarding timestamps is
always necessary.  This "clamping" concept... I have no idea where it
comes from.  It sounds terribly complicated.  And I don't think
kowtowing to the badly wrong (in the context of reproducibility, anyway,
certainly) cachebusting mechanisms of e.g. `make` has any place in a
well-crafted future.

And as PR material, overridingly, I agree that it makes sense to focus
on timestamps-are-variables please-declare-them.  Keep it simple. 
Pervasive awareness that build timestamps are a Bad Idea from the Last
Century has a lot of utility.  It's an easily remembered  rule that can
help new application developers sidestep making easy-to-avoid mistakes
in the first place.  Detailed awareness of rules for when to treat
application timestamp data like application timestamp data has very low
utility: it's A) almost always application specific and B) ought to be
fairly obvious to the application developer once they get in the weeds
and C) is hugely distracting to talk about up front.

I very frequently hand the reproducible-builds.org link to folks who
haven't heard of the concept before.  It does well as a brief
introduction on "Why is this important".   I like this: A good primer
for spreading awareness of the basics is way more impactful than an RFC,
at this point, IMO.

On Sun, Nov 6, 2016, at 02:29, Chris Lamb wrote:
> HW42 wrote:
> > Given that it's used in a bunch of cases I think it's reasonable to
> > include it in the spec.
> I'm in danger of sounding like a broken record here (!) but I strongly
> believe
> an overriding priority should be to be empathetic to the most common
> reader
> of the specification and thus not get too hung-up on being "technically"
> correct. :)
> As in, I fear "a bunch" might be misrepresenting the ratio between
> upstreams
> that might require clamping and those do not; in my experience the silent
> majority (that do not result in mails to this list) simply do not need to
> be
> worried with it, something in the order of 1000:1 or so.
> Don't forget that this quote-specification-unquote also serves as an
> advertisement for our endeavours.

More information about the rb-general mailing list