Re: [Gen-art] /.well-known placed below the URI local-part root

[Nico Williams <[e-mail]>]

On Fri, May 01, 2020 at 09:32:27AM +1000, Mark Nottingham wrote:
> (a) (b) Appendix A already talks about it some:
> 
> """
>    Why aren't per-directory well-known locations defined?
>       Allowing every URI path segment to have a well-known location
>       (e.g., "/images/.well-known/") would increase the risks of
>       colliding with a preexisting URI on a site, and generally these
>       solutions are found not to scale well because they're too
>       "chatty".
> """
> 
> We could say more in the spec proper, I suppose.

The more to say really is to have some examples showing how to do the
multi-tenancy thing.

I mean, it's obvious -- it's obvious to me anyways.

But evidently it's not actually that obvious.  At least one person I've
spoken to thought the text in section 3 of RFC 8615 does allow use below
the root, and I can see how such a misunderstanding can happen: "each
app has its own root, no?".  Because, yes, if you think about "mounting
an app" on the URI local-part "filesystem", then "each app has its own
root" makes sense.

> (c) seems like a good pattern to recommend for people who really want
> to scratch this itch -- although I suspect some will complain that
> it's hard to maintain the linkages (it's not really, they just need a
> competent sysadmin).

It's not an itch though, more like an accident.  People do this because
they build a service, they put it under /mysvc, then they add
multi-tenancy and don't set up a separate authority for each tenant
because hey, local-part components are cheap.

> (d) is not something I'm keen to do. Just because people have done it
> shouldn't mean that we should give that practice any formal air cover.

+1  (I wrote "possibly" because I'm not keen on it either.)

> (e) The current approach we take is that every registration is
> effectively its own name space; it can do what it likes with
> substructure (including establishing another registry). That gives a
> lot of freedom, but it's caused some confusion (e.g., most recently
> with EST and extensions to it), so some indication of what policy is
> in use would be interesting -- either in the spec, or the registry
> itself. It may be too early or this, but I could see a list of common
> management mechanisms being created for easy reuse.

Well, as you're the Expert Reviewer, maybe you can just start looking
out for this, asking registrants to write a blurb about this in their
specs and user-facing docs.  Maybe that's all that needs to be done for
the moment.

> Question: do you see any of these issues as urgent? .well-known got
> revised relatively recently, so I'd like to collect some more
> feedback/issues before reopening the box, ideally. If they're not
> urgent, would it be OK to just log them as open issues (I use
> <https://github.com/mnot/I-D/labels/well-known>)?

Not urgent.  I think this can be handled by the Expert Reviewer for new
registrations until more changes pile up for RFC 8615.

I've opened https://github.com/mnot/I-D/issues/315

Thanks for the quick reply!

Nico
--