Re: [apps-discuss] draft-ietf-weirds-bootstrap-00 and our lawn -- feedback?

["John R Levine" <[e-mail]>]

>> People keep saying this.  My client is a three line shell script that
>> uses wget and grep (really.)  Could you explain how that works with
>> templates?

We seem to be talking past each other here.

I believe that there are web servers that for reasons of bureacracy are 
run in ways with weird limitations, and it would be nice for them if http 
clients would do arbitrarily complex stuff to work around those servers' 
limitations.  Although it's hard to imagine why a domain or IP registry 
would use a server like that (none do now), it's not hard to imagine ISPs 
delegating the RDAP for an IPv6 /56 or /60 to a a SOHO router that's 
routing the IP traffic, where RDAP will share the tiny http server with 
the one for the config panel.

But there are also web clients that are rather constrained for both 
administrative and technical reasons, and "use templates" is not helpful 
advice.  (See unanswered question above.)

I also think I understand why it is not a good idea to invent random fixed 
URL syntax that people might shove into random places in a web server, but 
that's not what RDAP is proposing.  Each RDAP server picks its own 
arbitrary URL prefix which the bootstrap or upstream servers know about, 
and the RDAP stuff is all constrained to be under that prefix, not 
anywhere else in the name space. It's true, the syntax requires that some 
stuff be in the path and some as queries, but so be it.

As firmly as one side can say get better clients that can handle arbitrary 
templates, the other side can say get better servers that can handle the 
syntax that everyone uses.  Since there will be way more clients than 
servers, fixing the servers will minimize the global pain.

Having been through this kind of stuff before,* if RDAP is forced to stick 
in templates to get through the IESG, here's what will happen: a few 
clients that already have template libraries will use them.  Everyone else 
will see that the largest domain and IP registries use the syntax in the 
draft (their prototypes do now), and the small registries and 
subregistries will use the free python server commissioned by ICANN, which 
also uses the same syntax, so in practice you can skip the templates and 
it'll work.

A few registries or LIRs might take the spec at face value and use 
different URL syntax and expect the templates to deal with it. They will 
get a stream of complaints from people who tell them that their clients 
work fine with everyone else, you're broken, don't waste our time playing 
RFC lawyer.  So they'll eventually give up and stick in a rewriting proxy 
to match the defacto standard syntax, or for registries who are stubborn, 
helpful entrepreneurs will run proxies on their behalf which translate the 
queries, and also snoop on the query stream.  There are plenty of web 
WHOIS sites right now now that conveniently find the right WHOIS server 
for you and sell the queries to domain speculators, so this isn't a 
stretch at all.

I hope we agree that would be a ridiculous outcome.  If you want to help 
us, you need to understand RDAP enough to see what has a realistic chance 
of posing a problem in actual deployed implementations, and how to offer 
advice we can realistically follow.

R's,
John

* - I'm thinking of when SPF was forced to add a new RRTYPE