W3C home > Mailing lists > Public > www-tag@w3.org > March 2013

Re: media types as anti-pattern (was: Why polyglot is needed)

From: Eric J. Bowman <eric@bisonsystems.net>
Date: Mon, 25 Mar 2013 22:18:54 -0600
To: Harry Halpin <hhalpin@ibiblio.org>
Cc: www-tag@w3.org
Message-Id: <20130325221854.5a179adbb62ec67f68ea158e@bisonsystems.net>
Harry Halpin wrote:
> 
> Yet I'm of the "media types might be an anti-pattern" opinion the
> more I think about this given they can't be controlled often by
> document authors.
> 

An architecture not based on media types would change the nature of
intermediaries, obsoleting much of the deployed infrastructure of the
Web.  The result is some other pattern entirely, i.e. an architectural
style which is no longer REST, which has yet to be defined.

While problems do exist, the concept of media types is so central to
the Web architecture, that basically calling them a flaw in that
architecture drastically understates the ramifications -- deprecating
media types results in a completely undocumented and untested new
architectural style.

Scientifically, I have to be skeptical of this notion for a system as
massively peer-reviewed and deployed as the Web, when the proposed
mechanism for such sweeping change is a TAG edict not backed up by any
published academic work.

I believe that the concepts in REST are more important than the style
itself -- Roy devised a methodology for improving the Web architecture.
If, after this much passage of time, we come to falsify his conclusions,
hasn't the fantastic success of the modern Web vs. the original Web over
that time proven the methodology sound?

Falsification of REST's reliance on media types has not emerged as yet,
but if there's a case to be made (as some of you seem to believe), then
by all means make it, using Roy's proven methodology so we can all
follow your logic, and those qualified may participate in the peer-
review process involved in the publication of a scientific paper.

This paper may be made quite concise by stipulating to Dr. Fielding's
definitions and classifications (as they're the accepted science in the
field, now). The goal, starting with the null style, is to add
recognized constraints derived from existing styles, to create a new
hybrid style which addresses the limitations of the existing Web and
devises solutions to them.

Such a paper must then honestly compare and contrast the new style with
REST, in terms of scalability and security, based on prototype
implementations.  The TAG is tasked with the stewardship of the Web
architecture.  If the TAG determines that the architecture is outmoded
and needs changing, such a decision demands nothing less than getting
such an academic work out there for peer review *before* taking action.

Until then, I remain unconvinced that sniffing for magic numbers is any
more than a *different* solution, not a *better* solution, if I have to
leave the pros and cons to the unsubstantiated opinions of myself and
others.

-Eric
Received on Tuesday, 26 March 2013 04:19:10 GMT

This archive was generated by hypermail 2.3.1 : Tuesday, 26 March 2013 04:19:11 GMT