[rancid] Re: VLAN portion of rancid-run

[Saku Ytti]

On (2008-05-29 14:52 -0700), john heasley wrote:

> > > Adding 'terminal width' has concerned me; that it would reveal cisco/other
> > > platform bugs.  The number of platforms that I have is limited.  Who is
> > > using this change and with what platforms?
> > 
> > I'm using it on thousands of boxes (>5k or so), covering virtually
> > all devices running IOS and some running something similar enough
> > to support terminal width.
> > While of course I'd only need it on switching capable platforms,
> > but didn't bother making the distinction.
> 
> would setting it to something non-zero be a better choice?  128 for example.

I guess to answer this question, we'd need to take a peek at IOS source,
to see how wrapping is done, and would non-wrapping of infinitely large 
string force infinitely large buffer. I'd like to think that it's designer
lot better than that.
 What I do foresee with say 128 term len, is corner cases of people mailing
rancid-ml about why on some of their boxes they get weird diffs, making
it bit much harder to trace what's going on.

Anyhow the concern is valid, and as rancid is hugely popular and
typically upgraded without further testing internally, there is some pressure
to release software that does not crash all of your network at once.
 I'm not sure what is best way to address this, one way is to stagnate the
development 'we have what we need, and it works', one way is to just accept the
risk in new versions and try to warn users about it and yet another would be to
call for volunteers testing the changes in as heterogeneous networks as
possible, to give some level of trust on the change before introduced to
the wild.


-- 
  ++ytti