On Wed, Mar 1, 2017 at 1:41 PM, Patrick McManus
<pmcmanus(_at_)mozilla(_dot_)com>
wrote:
(2) Intermittent review of snapshots, ones that are generally
believed by their editors to be coherent and self-consistent.
Github revision tracking often makes that approach very
burdensome; change logs in documents and diffs between versions
are often more helpful.
This seems to suggest that github based drafts go from -00 to WGLC with
interim changes only happening on github. That's not the way it works in my
experience - to me it feels like the same number of drafts are produced
(some data here would be interesting.. http/2 had around 19 revisions at
the end iirc even though we used github extensively.. I believe tls 1.3 is
on draft 19 or 20 and it is also github centric) and they continue to have
changelogs.
Just to reinforce a couple of points mcmanus is making here:
1. I'm not seeing much evidence that people produce significantly fewer
draft
revisions when Github is in use. As patrick says, TLS is currently on -18
and
I'm about to do -19. To take three comparison points of roughly comparable
size:
- TLS 1.2 (RFC5246) was published after -09
- IKEv2 (RFC4306) was published after -17
- ICE (RFC 5245) was published after -19
My impression is that with Github, much like with non-Github, the draft
issuance
cadence is largely driven by the IETF meeting schedule. Certainly we've
found
that to be true for many of the drafts I am working on (JSEP and TLS in
particular), even though we use Github.
I'd certainly be interested in seeing some evidence that Github drafts come
out less frequently than ones that didn't use Github.
2. Changelog quaity varies, of course, and AFAIK there's no actual
requirement
to do a Changelog, but my experience is that the Github drafts actually do a
somewhat better job of Changelogs, probably because the tooling encourages
you to package things up in a way that makes it easy to do a ChangeLog. This
is especially true if you use PRs because then you can easily go back and
look at every merged PR and that becomes the input to the ChangeLog.
Here, for example, is the TLS ChangeLog.
https://tlswg.github.io/tls13-spec/#rfc.section.1.2
I think what *is* fair to say is that Github enables a far richer
interaction model
(Patrick alluded to this as well) with the result that it highlights the
relative
poverty of the old occasional draft model, but that's not because that model
has suddenly gotten absolutely worse, just that there are now better tools
available and people are taking advantage of them.
-Ekr