W3C home > Mailing lists > Public > www-tag@w3.org > January 2008

Re: Not UDP? Re: Proposed HTML ping attribute

From: Roy T. Fielding <fielding@gbiv.com>
Date: Thu, 17 Jan 2008 18:10:20 -0800
Message-Id: <5066C818-3341-4548-9A77-357A14EC2E6F@gbiv.com>
Cc: "Tim Berners-Lee" <timbl@w3.org>, www-tag@w3.org
To: Mark Baker <distobj@acm.org>

On Jan 17, 2008, at 2:13 PM, Mark Baker wrote:
> On 1/17/08, Tim Berners-Lee <timbl@w3.org> wrote:
>> The application semantics here are defined: it is a report of
>> a link-following action to a monitor.   People argue was to whether
>> it is more like an HTTP GET or HTTP POST, but whether the
>> ping protocol maps onto a GET or POST in HTTP is  question
>> you don't have to answer if you get a UDP port and define an
>> protocol spec for it.
>
> I don't understand.  The issue here, I think, is exactly whether the
> ping message should have GET or POST semantics (not that it'll
> necessarily have to use HTTP - I'm just referring to the meaning of
> GET and POST).  Whether TCP or UDP is used as a transport seems
> immaterial.  I certainly agree that UDP has advantages for that kind
> of message though.

I agree with Tim.  If the link traversal is indicated using a
protocol that is not HTTP, then HTTP semantics simply don't matter.
All that does matter is the safety of the protocol, which presumably
would be distinct from other UDP protocols to make it less likely
to be abused.

Note, however, that this is only a technical discussion of how one
might implement such a feature in a reasonably sane manner.  It does
not answer the more fundamental question of why HTML should have
this feature in the first place.

There is a very long list of reasons why websites use referral
services and redirects today.  Only one of those reasons (referral
detection) is satisfied by the ping proposal, and even then only if
nobody is allowed to turn it off and the monitor is willing to live
with counts that are far easier to defraud and won't detect
interrupted transfers.  I can't explain how these real systems
on real websites that people do almost all of their Internet
purchases upon work because I am legally obligated not to disclose.

I am not asking that people take my word for it.  I am asking that
the feature be vetted and justified by the professionals who are
expected to use this new "standard" before it becomes embedded as
yet another lame hack within what is supposed to be a mark-up
language.  If you can get LinkShare, Amazon.com, Yahoo!, and
Google (Marketing, not Engineering) to agree to use this feature
exclusive of redirects, then it should be standardized.

IMO, the only people who will use this ping attribute are the
web-juveniles who place self-incrementing dynamic counter images
on their personal home page.  This proposal creates way too high
a cost to satisfy such a small audience.

....Roy
Received on Friday, 18 January 2008 02:10:29 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Thursday, 26 April 2012 12:47:51 GMT