RE: Caching paragraph, #issue-html-nsdoc

[[
Transformation Caching
It can be useful for GRDDL aware agents to maintain a local cache of
material relating to GRDDL transformations to avoid unnecessary HTTP
requests. A typical case would be for namespace documents which rarely
change (and/or may not have an associated GRDDL transformation). This
caching may extend far beyond document expiry times provided by the
server. Where this is the case, care should be taken to ensure that the
cached information is current, to avoid systematic misinterpretation of
published data. Ideally implementers of GRDDL aware agents will allow
any local caching to be disabled. When caching  is not present, the
namespace document of a source document should be retrieved and checked
for the presence of GRDDL transformations. 
]] 
 
I think this is ok as is, but when I read it, it occurred to me that the
term "cache" suggests maintaining local copies of ns docs and the like
when all this is necessary is to retain a memory of the GRDDL transforms
indicated by those documents.  Also given DanC's experience of getting
locked out of the W3C site do we wish to encourage implementations to
maintain a cache.  If the W3C site cannot or does not wish to support a
very high rate of requests for common ns documents, perhaps we should
consider requiring some sort of memory.  These thoughts prompted to me
to suggest the following:
 
[[
Some standard namespace documents, such as the HTML [@@ref] namespace
document have very many references to them.  If GRDDL implementations
were to retrieve these documents every time they processed a document
referring to them, the servers serving those documents could become
overloaded.  GRDDL implementations therefore MUST NOT retrieve such
documents on every reference and MUST retain some local memory of the
transformations those documents indicate should be applied.  To avoid
misrepresentation of published information, GRDDL Implementations MUST
ensure that this local memory is up to date.
]] 
 
Brian