[ResourceTiming] Need to define "networking layer"

This is implementation feedback on Resource Timing.

The specification uses, without definition, the term "networking layer". 
  This term needs to be defined for the specification to be 
interoperably implementable.  Some questions that obviously arise:

1)  Is a file:// load coming from the "networking layer"?
2)  What differentiates a "networking layer" cache from other kinds
     of caches?  How do I tell whether a particular cache in a UA is
     a "networking layer" cache?  For example, Gecko has an in-memory
     image cache in its image library that largely implements HTTP
     caching semantics; is this a "networking layer" cache?  Note
     that this is _NOT_ the same cache as the cache the HTTP
     implementation uses.
3)  If a data: or blob: fetch requires the data to be read from disk
     because large data blobs are stored on disk, is that coming
     from the "networking layer"?  See question 1.
4)  If a data: or blob: fetch requires the data to be read over the
     network because on-device storage is too small to store
     significant data blobs, so they are pushed out to a different
     device for storage (whether explicitly or via file storage in a
     network filesystem), is that coming from the "networking layer"?

I can come up with more, but the upshot is that it seems like there are 
some assumptions being made about what this "networking layer" bit is, 
of the "I know it when I see it" variety, that need to be spelled out in 
the specification.

-Boris

Received on Thursday, 22 August 2013 02:58:44 UTC