Publishing code reviews

[Mark Reinhold]

> Date: Fri, 12 Oct 2007 11:10:35 +0100
> From: Andrew Haley <aph at redhat.com>

> Here's a really simple suggestion: convert the diffs to "diff -u"
> format and email them to a list.  ...

Your suggestions of sending diff -u output to mailing lists, for those
who prefer that format, and of making sure that those lists are archived
(in many places) are well taken.

It seems that there's a hierarchy of code-review formats in which each
format encompasses those below it.  The lowest level is simple uniform
diffs (diff -u), which are universal.  Next come webrevs, which include
uniform diffs, and then fancier, semi-automated systems such as the
"robot" (which already generates webrevs) and Review Board (which could
likely be hacked into doing so).

As long as we build infrastructure that can support every layer of the
hierarchy, everyone should be happy.

> ...
> 
> Finally, a server that only keeps webrevs for a limited period of time
> is a bad idea.  We need to be able to find every version of every
> patch that has ever been proposed.

In most cases it should be possible to reconstruct a webrev from the
corresponding diff in the e-mail archive.  The only exceptions I can
think of will be diffs that aren't relative to a specific published
Mercurial changeset.

Still, disk is cheap, and ZFS compresses HTML quite nicely, so perhaps
we should aim to keep webrevs around forever.

Thanks for your comments.  It's always good to get more outside
perspectives on this sort of thing.

- Mark