Re: Avoiding 'master' nomenclature

[Junio C Hamano <gitster@pobox.com>]

Johannes Schindelin <Johannes.Schindelin@gmx.de> writes:

> However, this might be overkill, and a bit more complicated to implement,
> as we now would also have to allow "negative" patterns.

Somebody may want to teach negative patterns to merge.suppressDest,
so the implementation complexity in the end would not be all that
different in the end, when that future happens.  But until then, I
tend to agree with you that it may be simpler if the matched ones
are suppressed than mentioned (iow, the configuration variable,
merge.suppressDest, would be simpler to manage than your
hypothetical merge.mentionDest whose polarity is opposite).

That is primarily because I expect that the common usage patterns
are the following three:

 - mention destination of merges into any and all branches;

 - mention destination of no merges;

 - mention destination of merges into all branches except for the
   primary integration branch.

A configuration variable with either polarity would express the
first two equally well, but the last one (which is the primary use
case for continuity reasons) is easier to express with suppressDest.

With mentionDest, you'd need two entries, i.e. 'all', and 'not this
one', so you'd need a negative matching from the get-go.  So from
the point of view of end-user usability, not ease-of-implementation,
I think Peff picked the right polarity in his suggestion.