Re: WS Description and "safety" [whenToUseGet-7]

On Tue, 2005-06-07 at 15:10 -0700, Jonathan Marsh wrote:
> Dear TAG,
> Almost a year ago, the TAG expressed some pleasure [1] at the WS
> Description WG's native support for operation safety.  As we have
> recently made some changes in this area, we thought it might be useful
> to provide an update to the TAG.
> Up through our recent public Working Drafts, we document a {safety}
> property and an associated attribute @safe as a property of the WSDL
> Core component model [2, esp. 6th bullet].  However, we received an
> issue [3] about the inability for tools to infer useful values for this
> attribute and the consequent impact on wide adoption.

Yes, well, I wouldn't expect that a tool could infer a useful value;
the whole point of our finding is that the designer has to *think*
about whether the operation is safe or not:

"we discuss the benefits of URI addressability, which is the primary
reason for distinguishing GET from POST. We then characterize safe
interactions, when the use of GET is encouraged. However, when
considering GET or POST, designers should also remain aware of
considerations for sensitive data such as passwords or credit card
information, and other practical considerations."

>   This discussion
> exposed significant differences remaining in the Working Group around
> operation safety.  We had objectors to removing the property altogether,
> yet also had significant objection to retaining the status quo.
> We settled on a compromise in two parts:
> 1) The "safe" attribute would be moved to an extension namespace and
> documented in the Adjuncts spec instead of the Core spec.  This
> satisfied those who felt that a semantic extension was inappropriate in
> the Core (which otherwise concerns itself with the mechanics of
> communication, not the semantics.)

That doesn't seem like a good idea. I'll have to think some more
about whether this compromise is acceptable.

> 2) The HTTP binding would take a dependency upon the safe attribute, and
> when the safe property is set to true, the HTTP method would default to
> GET, though this can be overridden when other considerations (e.g. data
> not easily serialized in a URI) apply.  This satisfied those who felt
> safety was important enough to not only remain in the family of
> Recommendations, but to have a real effect on the HTTP binding.
> I took an action as chair to update you on the current situation.  I'd
> be happy to answer questions or provide you more detail if you wish.

Thanks for keeping us informed.

> - Jonathan Marsh
>   WS Description WG
> [1]
> [2]
> [3]
Dan Connolly, W3C
D3C2 887B 0F92 6005 C541  0875 0F91 96DE 6E52 C29E

Received on Wednesday, 8 June 2005 17:19:32 UTC