On Tue, Jun 15, 2021 at 12:03 PM Paul Wouters <[email protected]> wrote:
> On Tue, 15 Jun 2021, Shumon Huque wrote: > > > On Tue, Jun 15, 2021 at 12:46 PM Tim Wicinski <[email protected]> > wrote: > > > > Yes, Stephane, we were envisioning recommending an > underscore label. Of course, that leads to how to avoid collisions in that > > space, and whether we need to establish a registry of > application service names. > > > > > > You mean, a different registry than this one > > > https://www.iana.org/assignments/dns-parameters/dns-parameters.xhtml#underscored-globally-scoped-dns-node-names > > > > tim > > > > > > Tim - yes, I think this would be a bit different. The above is for IETF > defined protocols. This one (if we think it's a good idea) would have to > encompass > > arbitrary Internet application services, many that could be proprietary > services of companies. > > <hat> I am one of the underscored-globally-scoped-dns-node-names Experts > > The _underscore registry is "Expert Review" only, meaning it is not only > used for IETF defined protocols. It's only goal is to be a place where > people can register a unique name to avoid name collision between > different protocols/applications using it. > > As such, it would be fine for this draft to commend registration there. > It could also start its own _underscore registry. > > </hat> > > Of course, if people ensure the names they use are somehow linked to > their product of business name, it becomes fairly unique to begin > with, and a registry might not be needed. Like people shouldn't be > using _registration or _website_auth or something generic like that. > My personal preference would be to focus stronger on generating proper > names (and embedded expire / recurring check within the name) that > would ensure no central registry of any kind would be needed. > The only real issue I see is managing the underscore namespace as a flat namespace. Given that DNS itself was engineered in part to be hierarchical as a solution to the scaling problems of hosts.txt, maybe using an underscore scheme that is hierarchical would solve some/many of these problems? (Perhaps carved out with its own underscore suffix.) The analogy that comes to mind is the MIB tree, with the public branch (.1.3.6.1.2.1) and the enterprise (.1.3.6.1.4.1). The enterprise portion of the tree has a very large number of nodes. Using numeric values rather than names avoids name collisions. These are just suggestions though. YMMV. Brian
_______________________________________________ DNSOP mailing list [email protected] https://www.ietf.org/mailman/listinfo/dnsop
