Re: Suggest slight change in description of Slug

[Bill de hOra <bill@xxxxxxxxxx>]

Thomas Broyer wrote:

> Slug was defined as a request-header in PaceSlugHeader2 [3] but
> PaceSlugHeader4 [3] finally made it in draft -10, where it's defined
> as an entity-header. See also PaceSlugHeader3 [5].
> The "when accompanying a POST" thing appeared in draft -12. 

On the whole, this was never a well-defined feature. The change in 
draft-12 made the use of the header sane wrt the protcol actions. Would 
you prefer we said nothing?


> The "short
> name" thing disappeared in draft -14. 
> [...]
> And please, editors, bring back the "short name" thing.

draft-13:
"Slug is a HTTP entity-header whose value is a short name that, when 
accompanying a POST to a Collection, constitutes a request by the client 
that its value be used as part of the URI for the to-be-created Member 
Resource."


draft-14:
"Slug is a HTTP entity-header that when accompanying a POST to a 
Collection, constitutes a request by the client that its value be used 
as part of the URI for the to-be-created Member Resource."

"short name" had no meaning. I'm very happy to take out such 
terminology.  I expect it might have been some latent reference to 
atom:title, but who knows.

If you can crisply explain what "short name" meant, and why not having 
it has impact to the protocol, I'm prepared to listen.


> This is, I think, mainly due to
> the fact that the editors never were in favor of either the Slug
> header in itself or it being an entity-header.

Rubbish. This editor had two main concerns:

 - SLUG appropriates a term used in blogging, but whose mechanism was 
not implemented in any blog software, until the i18n requirement was 
clarified. be aware that "Servers SHOULD treat the slug as [RFC2047] 
encoded" largely exists so that stacks can choose to support the feature 
without reimplementing their slugging code for i18n.

 - SLUG is untestable without re-implementing what each server does; 
Tim is about to find this out, as will others. The "slight change" is an 
attempt to insert precision in the format data, but generally it's a 
failure to admit that the best you can do with the feature is to reverse 
engineer what servers do.

cheers
Bill