Re: [apps-discuss] RFC 5785: Registration of .well-known services under HTTP to First Come

["Phillip Hallam-Baker" <[e-mail]>]

Roy's response accuses me of lying, I regard that as defamatory.

 

 

 

From: apps-discuss [mailto:[e-mail]] On Behalf Of Roy
T. Fielding
Sent: Tuesday, January 19, 2016 11:43 AM
To: Phillip Hallam-Baker <[e-mail]>
Cc: General discussion of application-layer protocols
<[e-mail]>
Subject: Re: [apps-discuss] RFC 5785: Registration of .well-known services
under HTTP to First Come

 

On Jan 16, 2016, at 10:48 AM, Phillip Hallam-Baker <[e-mail]
<mailto:[e-mail]> > wrote:

[Matthew Kerwin wrote:]

wasn't .well-known acknowledged to be an anti-pattern when it was set up?
And isn't the DE role therefore to turn proposals away which would be better
served by a (positive) pattern elsewhere, along with that explanation?

The bar for .well-known should be very high because http != gopher, nor
should it ever be.

No, it was asserted but never conceded. More importantly, the IETF reviewed
and agreed the RFC which makes no mention of such a position. 

 

http://tools.ietf.org/html/rfc5785#section-1.1





The position of DE is not meant to be an opportunity to impose a private
view of Internet architecture on anyone. It certainly isn't an opportunity
to tell people that they should 'do it my way' and especially not when that
turns out to be implement someone's 20 year old thesis.

 

The .well-known namespace has nothing whatsoever to do with REST.  It is
based on the

general idea originally exposed by /robots.txt: a set of well known
resources for which no link

is necessary, such that a client can attempt to perform some defined
(standard) prerequisite

action before some other "normal" action on that same origin server.

 

http://www.iana.org/assignments/well-known-uris/well-known-uris.xhtml





I am well aware of the style of architecture that people are suggesting it
is the role of the DE to impose. It is a style that I played as much of a
part in creating as anyone else did. It is a style that I have used and have
since abandoned because I think other approaches are better. 

 

You have not made a registration request of the DE, nor has anyone suggested
that your

poor use of DNS, SRV, and .well-known had anything to do with Web Services
vs REST.

Darrel Miller disagreed with your characterization of caching and use of
HTTP as an

application tunnel, which is what you should expect when making statements
on a mailing

list like apps-discuss that directly contradict all known experience with an
IETF protocol.

 

https://mailarchive.ietf.org/arch/search/?email_list=wellknown-uri-review

 

Mark suggested here that you need to pick a name to register, in accordance
with the RFC.

 

https://mailarchive.ietf.org/arch/msg/apps-discuss/CACiBdiDnlW1X7JsnXdEXW22m
1U

https://mailarchive.ietf.org/arch/msg/apps-discuss/X_MTE2a4xKeaPh2zOSlt4qmjN
mg 

 

I suggested here that you can register a name even if the DE disagrees with
your usage,

as there is plenty of actual evidence of that in the mailing list, that a
registration discussion

is helpful for preventing misinformed folks from cluttering up the space
when they actually

want something else (usually a Link header field), and that such discussion
is also useful

for providing a historical record of why registrations are made regardless
of suitability.

 

https://mailarchive.ietf.org/arch/msg/apps-discuss/s0JohIW0qcr7ajHUDyJD4bJP8
6U

 

The purpose of a DE is to ensure that the registrations meet the criteria of
the namespace

as defined by the relevant RFC. Equating all namespaces with port numbers is
nonsense.

Port numbers cannot contain trademarks, offensive slogans, or blatant
categorical errors.

The .well-known space would be misused under a first-come first-served
basis.

 

And, no, SRV would not have been used by "http" in 1992, nor would it be
used in

2016, since having multiple independent Web servers per host has always been
a desired

feature (to escape the largely MIS-controlled gopher phenomenon) and it is
still common

today to identify configuration websites by local host names and IP
addresses.

The MX record as architecture is ideal for the social characteristics of
email hosting,

not for the comparatively anarchic nature of website deployment.

 

Since then, you have repeatedly misquoted and misrepresented what we said in
your own

self-serving arguments. If anyone wants to discuss the larger role of DEs
within the IETF,

I hope that they do so based on actual events and not on your fabrications.

 

....Roy