W3C home > Mailing lists > Public > www-ws-arch@w3.org > July 2002

RE: [rest-discuss] RE: Clarification on REST, GET and CGI

From: David Orchard <dorchard@bea.com>
Date: Thu, 25 Jul 2002 09:50:42 -0700
To: "'Bill de hÓra'" <dehora@eircom.net>, "'Paul Prescod'" <paul@prescod.net>
Cc: <www-ws-arch@w3.org>, <rest-discuss@yahoogroups.com>, <fielding@apache.org>
Message-ID: <04ee01c2340d$d77b33b0$0100007f@beasys.com>

Bill,

Very interesting suggestion, thank you for taking the time to write it.  I
take it from your comments that you are at least pleased that we, the TAG,
can write architecture in the positive sense.  One of the mechanisms we have
been using is the difference between the architecture document and the
findings.  We are trying to put "the positive" into the architecture doc,
and have the extended discussion in the findings.  I think this goes partway
to what you are looking for, but not as far as you'd like.  I wonder how we
could get an ongoing, useful and open discussion going.  Thinking aloud
(electronically so to speak), maybe we could have TAG findings refer to well
written patterns or anti-patterns..  At any rate, the TAG is certainly open
to ideas on how to better communicate with the various constituents, and how
to receive feedback/input.  Much of a successful architect's job is
communicating.

FWIW, I believe that we have a real disconnect and inability to talk about
various architectural styles in a detailed and positive manner.  I know I'll
get into trouble saying this on the REST discussion list, but none of us are
really shrinking violets...

Any architecture, REST or otherwise, makes certain trade-offs that hinder
some users or agents, and help others.  There may be a REST anti-pattern -
such as methods in the SOAP body anybody? - that are patterns in another
architecture.  Now there are trade-offs to either.  One of Roy's points
about why methods are bad in the body is because it adds another element to
the security context, particularly encumbering the firewall admin.  On the
ws-arch list, I wrote up 3 various ways that a simple GET/SET stockquote
could be done using 1) method names in body, 2) GET/POST; 3) different URLs.
And then I drilled into the way the firewall admin would administer each of
the designs.  This unfortunately didn't spark much discussion.  When I've
spent time with others - such as Mark Baker - and tried to drill into this
issue, it's not clear to me how significant the down side is.  So I admit
that even though I can quote scripture - even wrote up some REST/shared
information space scripture for the TAG bible - I haven't gotten the "faith"
that there is only one acceptable architecture for what we want to do with
much of machine to machine communications.

What I would ask from the REST community is that it be open to rationale,
detailed and reasoned discussion about the trade-offs associated with
diverse architectural styles.  Too much of the debate so far has been simply
n party broadcast monologues, to paraphrase MLK.

Cheers,
Dave


> -----Original Message-----
> From: www-ws-arch-request@w3.org [mailto:www-ws-arch-request@w3.org]On
> Behalf Of Bill de hÓra
> Sent: Thursday, July 25, 2002 12:25 AM
> To: 'David Orchard'; 'Paul Prescod'
> Cc: www-ws-arch@w3.org; rest-discuss@yahoogroups.com;
> fielding@apache.org
> Subject: RE: [rest-discuss] RE: Clarification on REST, GET and CGI
>
>
>
>
>
> > From: David Orchard [mailto:dorchard@bea.com]
> >
> > hmm.  Roy said that many CGI scripts suck because they don't
> > implement REST properly.  Don't think I'm misquoting him.
> > I'm pointing out that we don't look at the sucky CGI scripts
> > when determine web architecture because, well, they suck.
>
> David,
>
> Yes, on the other hand worst practices and antipatterns are highly
> valuable guides to developers in the trenches, especially given the
> amount of trench and the way people tend to cut and paste web systems
> together (CGI being a prime example). For example, saying
> it's better to
> not invent a new URI scheme is not quite the same as saying
> an arbitrary
> new scheme sucks (cue: reason). As one example, the OO world
> has learned
> much via antipatterns. Today, I suggest that saying non RESTful CGIs
> suck is less useful that concrete examples; REST is not exactly
> mainstream thinking yet.
>
> There's a good amount of physical architecture in the world that is
> based on what not to do as much as what to do. One thing the TAG could
> say to the web community is, look we can't do it all, you guys start
> writing the anti-patterns down and let us concentrate on the
> architecture in a positive form.
>
> regards,
> Bill de hÓra
>
> ..
> Propylon
> www.propylon.com
>
>
Received on Thursday, 25 July 2002 15:04:20 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Tuesday, 3 July 2007 12:25:03 GMT