Re: [I18nrp] Additional input needed for i18nRP BOF

[Adam Roach <[e-mail]>]

On 6/6/18 9:45 PM, Nico Williams wrote:
> On Wed, Jun 06, 2018 at 05:56:47PM -0500, Adam Roach wrote:
>> On 6/6/18 5:31 PM, Peter Saint-Andre wrote:
>>> SECDIR, GENART, and ARTART reviewers seem to check the whole document.
>>> By contrast, I18N reviews would have a smaller scope (along the lines of
>>> what the SDPDIR does).
> FWIW, I concur with Peter that your message about directorate
> organization and functions was quite useful.
>
>> I'm not sure how good a comparison this is. Any document that doesn't
>> mention SDP is clearly not a candidate for SDPDIR review. Any document that
>> doesn't mention i18n, probably *does* need a check that such omission makes
>> sense. In that way, I think the required task is more like SECDIR than
>> SDPDIR.
> Might it be reasonable to expect shepherds and/or responsible ADs to
> detect I18NDIR review opportunities and refer documents to it?

Perhaps, but my experience is that such things can easily fall between 
the cracks when left to shepherds and ADs.  Possibly, we could amend the 
shepherd writeup template to ask whether the document deals at all with 
human-readable text; and, if so, whether a directorate review has been 
requested -- perhaps that would work. I do note that the shepherd 
write-up is already quite long; and that (presumably as a consequence) 
such questions are often given pro-forma responses. On the occasions 
that I run id-nits on documents submitted to the IESG for review and 
find issues, I've found that about half of the shepherd's writeups 
answered the question about id-nits with the incorrect statement that no 
issues were found.

So while it would probably be ideal to rely on AD and/or shepherds to 
flag documents, I suspect we'll get better coverage by having a 
dedicated directorate triage (with an obvious increase in load on the 
directorate). I'm not sure what the best trade-off is here. Perhaps we 
run a series of experiments to see what works best.

> Regardless of what the answer to that is, it might be possible to start
> small and scale up.  Starting small means triaging more demurely.  At
> any rate, starting small is probably the only realistic way to get
> started.  But there's no point starting small if we can't scale up, so
> we do need to have some idea of how big to scale, and how.

This sounds reasonable, and is pretty much what I had in mind when I 
said "volunteers who are willing and able to educate themselves into 
being experts". I say this while agreeing with John about the 
complexities and subtleties of this topic.

/a