Requesting review of "Eliminate some pointless gymnastics in the config parser."
Eric S. Raymond
esr at thyrsus.com
Fri Jul 8 12:35:23 UTC 2016
Frank Nicholas <frank at nicholasfamilycentral.com>:
> > On Jul 8, 2016, at 12:28 AM, Eric S. Raymond <esr at thyrsus.com> wrote:
> > (1) Sometimes the item would directly set a field in the new node. This
> > is the case with ttl and minpoll.
> > (2) For flag items, such as iburst and xleave, the iterm would get put
> > on a newly created list attached to the node. Later, when some other
> > function wanted the value of the flag mask for this entry, it would
> > call a helper that would walk that FIFO and | the values together.
> > What's puzzling me is: why not just | all those flags together at the
> > time of the initial parse, and stash that in a flag field? Why create the
> > linked list and delay evaluation?
> Would this be related to the “interleaved” mode that waited for the next round of data to complete some items? I don’t remember the details, but it should like it was waiting for two sets of data.
No connection. The config parse ends before protocol exchanges are active.
<a href="http://www.catb.org/~esr/">Eric S. Raymond</a>
More information about the devel