On 17/11/2013 3:21 p.m., Alex Rousskov wrote: > On 11/15/2013 11:49 PM, Amos Jeffries wrote: >>>> Now about the name: "realm_format" is a bad choice IMO because some >>>> folks will think that it controls the format of the authentication realm >>>> string displayed to the user (for schemes where we can specify that >>>> user-visible string). I suggest calling the new option "request_extras". >>>> The configured extras will be appended to the helper request and to the >>>> cache key. Any better naming ideas? > >> Hmm. We could call it "notes" or "annotations" and document it as part >> of the custom annotations and other details being sent to the helper. > > > I do not like "notes" or "annotations" because, in all other contexts > already in use, those things denote information added by an "external" > force such as an admin, helper, or adaptation service. In the case of > "request_format", these additional details would usually be a part of > the transaction already (and may eventually include other annotations!). > > Another reason to reject "notes" or "annotations" is to avoid the > implication that their use annotates the affected transaction as if a > "note" option was used (or as if an adaptation service created an > annotation). > > Finally, all existing annotations are key=value pairs while these new > details do not have such a well-defined structure. They can be viewed as > one big anonymous annotation, but it is probably a bad idea to create > such anonymous annotations. >
Okay. > > Please note that the context of the new name is "authentication > parameter" as already determined by the auth_param directive itself: > > auth_param digest program ... > auth_param digest children 20 startup=0 idle=1 > auth_param digest realm Squid proxy-caching web server > auth_param digest request_extras "%lp" > > Request_extras is not ideal because it does not explicitly tell the > amdin that the new parameter affects the authentication cache indexing > (the documentation will say that, of course). We can use something like > key_suffix. It would not explicitly tell the admin that the helper > request format includes that suffix, but it is not worse than > request_extras IMO: > > > auth_param digest key_suffix "%lp" Halfway: key_extras ? Amos
