On 10/12/2012 07:00 PM, Alex Rousskov wrote: > OK, so here is the summary of the required changes for the proposed > meta_header patch: > > 1. Rename proposed "meta_header" directive to "note". > 2. Rename proposed "%meta_header" logformat code to "%note". > 3. Enforce restrictions on characters in keys if not already > enforced. > > Going forward, if somebody wants to support "stacked tags" using > external ACLs, they will: > > 1. Add a "note=key:value" keyword support to external ACLs > 2. Add a new "note" ACL type that will match key:value annotations. > 3. Adjust external ACL code to store tags as tag:<value> annotations, > using annotations API added by the meta_headers patch. > 4. Adjust "tag" ACL type code to look for tag:<value> annotations, > using annotations API added by the meta_headers patch. > 5. Make sure that external ACL annotations are not sent to adaptation > services (enabling such sending may be a separate project). > > Any objections to this plan?
Just some comments. Currently the meta_headers as implemented in the patch, they share the same code with adaptation meta headers. They are more or less the same thing and have similar usage. In the case of adaptation the meta headers are send to the service as headers (so they are headers). They are headers which are used to pass meta data to the service. This patch starts adding a similar mechanism inside squid: Add support for internal headers used to pass meta data to various squid subsystems. The tag/note notation looks a little confusing to me, and I am not sure I am understanding why they are better ... squid is an HTTP proxy, and the HTTP protocol uses headers to pass control/meta/info data. So it looks less confusing to me to just add in squid a mechanism which have the meaning of "internal headers". The system admin needs only to know that these headers are only for internal use. Just some thoughts, no need to pay attention... > > > Thank you, > > Alex. > >
