Re: [netmod] Pattern statements

[Martin Bjorklund <[e-mail]>]

Ladislav Lhotka <[e-mail]> wrote:
> t.petch píše v St 23. 08. 2017 v 17:28 +0100:
> > ----- Original Message -----
> > From: "Ladislav Lhotka" <[e-mail]>
> > Sent: Wednesday, August 23, 2017 11:53 AM
> > 
> > > "t.petch" <[e-mail]> writes:

[...]

> > > > time in base YANG before anyone noticed - it was just too complex.
> > > 
> > > Why was it wrong? Just because it was too complex?
> > 
> > No; it contained a definite error.
> > 
> > This was probably in yang-types and probably around 2012, quite late in
> > the day, and it stuck in my mind that so many had looked at it and
> > failed to spot that it was wrong, not just that it did not cater for
> > some aspects such as interface I-D.  I have used it before as an example
> > of over complexity
> > 
> > I will have the e-mail filed, along with several thousand other NETMOD
> > ones so I will find it later rather than sooner.
> 
> I believe you mean the case when it was realized that "ipv4-address" and "ipv6-
> address" permit also zone indices: 
> 
> https://www.ietf.org/mail-archive/web/netmod/current/msg07456.html
> 
> We can hardly blaim the pattern expression for this though.
> 
> Other than that, I am not aware of any reported problem concerning the "ipv6-
> address" type. In fact, I even don't think the two ANDed regexs are excessively
> complex. Just compare them to ABNF productions proposed for the same purpose in
> RFC 3986, Appendix A.
> 
> Lada
> 
> PS. Yes, it's my child, so I do have a reason to feel offended. :-)

Note that we had a different pattern to start with.

The new pattern was added 2009-03-27, and it is still the same as in
RFC 6991.

The new pattern was a simplified version using AND; the old version
was a single pattern.


/martin