On Thu, Jul 13, 2017 at 02:49:28PM -0700, Stefan Beller wrote:
> Snarkily I wanted to link to an essay "large patch series
> considered harmful"[1], but could not find any. So a couple
> of bullet points:
> (a) humans dislike larger series, hence fewer reviewers
> (b) the likelihood of a mistake in new code is proportional to its size
>   We can use the number of patches as a proxy for size
> (c) If a mistake is found, the whole series needs rerolling.
>   The effort of rerolling a series can be approximated with
>   its size as well.
> 
> From combining (b) and (c), we conclude that the effort to
> land a patch series is O(n^2) with n as the number of patches.
> Also from (a) we conclude that two smaller series containing
> the same output as one large series, has better code quality.
> So with that, we conclude that all series shall be as small
> as possible.
> 
> So I'd ask to queue these 2 separately, asking Brian to drop
> "builtin/verify-tag: convert to struct object_id"
> 
> [1] https://www.cs.princeton.edu/~dpw/papers/hotos.pdf, 2005
>     seems interesting to me in hindsight.
> 
> I can also send my patches to Brian, as you (both) like.

I'm literally about to send my series out; I rebased and tested it last
night.  I don't care whose patch gets applied, but to avoid needing to
rebase and retest my series, I'm going to send it out as it is.  Junio
can apply my series on top of yours, in which case he can drop the
relevant patch from my series.
-- 
brian m. carlson / brian with sandals: Houston, Texas, US
https://www.crustytoothpaste.net/~bmc | My opinion only
OpenPGP: https://keybase.io/bk2204